2021 enterprise trends

in machine learning
Contents
Introduction .................................................................................................................................................................................................... .01

Survey at a glance ..................................................................................................................................................................................... .02

Report theme 1: Organizations are increasing AI/ML budgets, staff, and

number of use cases—but the market’s still dominated by early adopters ...................................................... .04

Report theme 2: Organizations continue to face challenges

across the ML lifecycle, and struggle with cross-functional alignment ............................................................... .13

Report theme 3: Despite increased budgets and hiring, organizations

are spending more time and resources—not less—on model deployment ...................................................... .23

Report theme 4: With increasingly complex infrastructure needs,

organizations report improved outcomes with third-party MLOps solutions ................................................. .27

Conclusion ....................................................................................................................................................................................................... .32

Methodology .................................................................................................................................................................................................. .34

About Algorithmia ..................................................................................................................................................................................... .37

Introduction
2020 was a year of belt-tightening for many organizations due largely to the macroeconomic
impacts of the COVID-19 pandemic. In May 2020, Gartner predicted that global IT spending
would decline 8% over the course of 2020 as business and technology leaders refocused
their budgets on their most important initiatives.

One thing is readily apparent in the 2021 edition of our enterprise trends in machine
learning report: AI and ML initiatives are clearly on the priority list in many organizations.
Not only has the upheaval of 2020 not impeded AI/ML efforts that were already underway,
but it appears to have accelerated those projects as well as new initiatives.

This is our third year conducting a survey about the current state of enterprise machine
learning and reporting on the results, and this is our most comprehensive report yet.
This year we expanded our questions about organizational maturity, a critical indicator
of long-term success that includes organizational alignment, data, training, deployment,
management, and governance. To improve the quality of our data and allow more in-depth
insights, we also increased the specificity of several questions and the options given to
respondents for answering them.

This report is the result of that effort, and includes our detailed analysis of the findings.
It is clearly an exciting—and exceedingly important—point in the AI/ML journeys of many
organizations. Algorithmia is committed to being good stewards of machine learning
technology and critical partners in our customers’ success.

1
“
Machine learning operations is a nascent discipline; it isn’t well recognized in the
industry. The survey results in this report underscore the importance of MLOps and its
impact on optimizing infrastructure consumption, improving application performance, and
increasing productivity for data scientists. As ML has come to the mainstream, it’s become
apparent that MLOps is the key to driving efficiency and scale for organizations of any size.
ML is both an important and emerging category that C-level leaders will prioritize
if they want to stay competitive.”

Tim Tully
CTO and SVP, Splunk

Survey at a glance
This year’s survey revealed 10 key trends that organizations should be paying attention to
if they want to succeed with AI/ML in 2021. The trends fall into a few main themes, and the
overarching takeaway is that organizations are moving AI/ML initiatives up their strategic
priority lists—and accelerating their spending and hiring in the process.

But despite increasing budgets and staff, organizations continue to face significant barriers
to reaping AI/ML’s full benefits. Specifically, the market is still dominated by early adopters,
and organizations continue to struggle with basic deployment and organizational challenges.
The bottom line is, organizations simply haven’t learned how to translate increasing
investments into efficiency and scale.

However, we remain optimistic about the growing potential of AI/ML in 2021 and beyond.
As the space continues to mature, the barrier to entry has continued to get lower. The
time to invest in AI/ML is now—no matter your organization’s size, industry, or unique
infrastructure needs.

2
2021 top 01 02
10 trends in
enterprise ML Priority and budgets
for AI/ML are increasing
Organizations are
expanding into a wider
significantly year-on-year range of AI/ML use cases,
with particular focus on
process automation and
customer experience

03 04 05
Most organizations have Governance is by far The second greatest
more than 25 models in the top challenge for AI/ML challenge is
production, but there’s AI/ML deployment, with technology integration and
a gap between AI/ML more than half of all compatibility, with 49% of
“haves” and “have-nots” organizations ranking organizations ranking it as
it as a concern a concern

06 07 08
Successful AI/ML Organizational The time required
initiatives require alignment is the biggest to deploy a model is
organizational alignment gap in achieving AI/ML increasing, with 64% of
across multiple maturity all organizations taking a
decision-makers and month or longer
business functions

09 10
38% of organizations Organizations that Continue reading to
spend more than 50% use a third-party explore each of these
of their data scientists’ machine learning 10 trends in detail,
time on deployment—and operations solution and learn how your
that only gets worse save money and spend organization can make
with scale less time on model the most of your AI/ML
deployment than investments in 2021.
those that build
their own solution
 Report theme 1: Organizations are
increasing AI/ML budgets, staff, and number
of use cases—but the market’s still dominated
by early adopters
It’s clear from this year’s data that AI/ML projects have become
The time to invest in one of the top strategic priorities in many enterprises. As of last year,
AI/ML is now, no matter organizations had already begun to boost their AI/ML investments;
your organization’s size, 71% of respondents in our 2020 report said their AI/ML budgets had
industry, or infrastructure increased compared with the previous year.
needs. They’re not dialing back that spending this year. In fact, companies
appear to be doubling down on their AI/ML investments. We ran a
survey this summer to see how organizations were adapting to the
pandemic and its impacts, and it showed a new sense of urgency
around AI/ML projects.

When we asked respondents why, 43% said their AI/ML initiatives

“matter way more than we thought.” Nearly one in four said that their
AI/ML initiatives should have been their top priority sooner.

It’s not just talk: They’re now ramping up their budgets and staffing

43 % accordingly. In our summer survey, 50% of respondents indicated that

they plan to spend more on AI/ML this year. Roughly one in five said
Respondents who told us they “plan to spend a lot more”.
that AI/ML matters “way
That’s reflected in AI/ML-related hiring, too. At a time when many
more than we thought” organizations are making difficult decisions about staffing, 76% of
in a survey this summer respondents in our summer survey said they had not reduced the size
of their AI/ML teams—with a full 27% reporting they had increased it.

We see clear evidence in our 2021 survey data that these trends
are expanding and extending in the year ahead in spite of continued
economic concerns and other uncertainty. In fact, those economic
concerns themselves have had a catalytic effect: Organizations are
developing a laser focus on AI/ML initiatives as a means of driving
top-line revenue while keeping bottom-line costs under control to
ensure they are as competitive as possible.

Trend 1: Priority and budgets for AI/ML are increasing

significantly year-on-year
Continuing the trends we saw in our summer survey, our 2021 survey
shows an increase in prioritization, spending, and hiring for AI/ML. First
off, 76% of organizations say they prioritize AI/ML over other IT initiatives,
and 64% say the priority of AI/ML has increased relative to other IT
initiatives in the last 12 months.

4
76% of organizations prioritize AI/ML over other IT initiatives

4%
Low priority

20%
Medium priority

49%
High priority

28%
Top priority

Respondents were asked to indicate the priority given to AI/ML initiatives relative to other IT initiatives at their organizations. Those
who said they were prioritizing AI/ML as either top or high priority relative to other IT initiatives were considered to be prioritizing
AI/ML over other IT initiatives. The total percentage of these respondents was calculated with the underlying data before being
rounded to the nearest percentage point. Categories do not add up to 100% because they have been rounded to the nearest
percentage point.

64% of organizations have increased AI/ML priority in the past year

6%
Decreased priority

31%
Same priority

64 %
Increased priority

Respondents were asked to indicate how the priority of their AI/ML initiatives has changed relative to other IT initiatives in the past
12 months. Categories do not add up to 100% because they have been rounded to the nearest percentage point.

5
When it comes to spending, a full 83% of organizations have increased their budgets year-on-year.

83% of organizations have increased AI/ML budgets year-on-year

Budget change for FY 2018-19 Budget change for FY 2019-20
50%

43%

40%

34%

30%
29%
27%

21%
20%

13% 13%

10%
7%
4% 3% 4%
2%
0%
1-25% 26-50% 51-75% <75% No change Decrease 1-25% 26-50% 51-75% <75% No change Decrease

Increase Increase

71% 2% 83% 4%
Increased budget Decreased budget Increased budget Decreased budget

13%
27 % No change
No change

FY 2018-19 FY 2019-20

6
You might think that this is ordinary growth attributable to the general growth of AI/ML. There’s partial truth
to this. However, it doesn’t account for the reality that 2020 has been a year of extraordinary change and
upset in the industry, a year in which many companies are looking for opportunities to scale back all but
their most mission-critical priorities. It also doesn’t tell the full story in terms of how much organizations are
increasing their AI/ML budgets.

In last year’s report, just 7% of respondents said their budgets were increasing by more than 50%. This year,
20% reported AI/ML budget increases of more than 50% from FY2019 to FY2020.

Here’s an even more telling indicator of the accelerating pace of AI/ML strategies. Respondents were asked
how many data scientists their organizations employ, from which we estimated the average number of data
scientists employed by organizations in both the 2020 and 2021 data. Year-on-year, the average number of
data scientists employed has increased by 76%. In fact 29% of respondents in our 2021 report now have
more than 100 data scientists on their team, a significant increase from the 17% reported last year.

The average number of data scientists employed has increased

76% year-on-year
2020 2021
50 %
60

50
40% 50
40%
+76%
difference

40

30 % 29% 29%

24% 30 28

20%
17%
15% 15% 14% 20
13%

10%
10
5%

0% 0

<10 10-19 20-49 50-100 >100 <10 10-19 20-49 50-100 >100 Avg. data
scientists

In 2020, respondents were asked to indicate the number of data scientists employed at their organizations through a free-form
response field. In 2021, respondents were instead asked to indicate the number of data scientists employed based on predefined
ranges, such as “20-49”. To create an accurate year-on-year comparison, 2020 data was first aggregated into the same predefined
ranges as 2021. Then, the average number of data scientists employed was estimated by using the lower bound of data scientists
in each group (for example, 20 for “20-49”) and averaging across respondents. The percent difference was calculated with the
underlying data before rounding to the nearest percentage point. Categories do not add up to 100% because they have been
rounded to the nearest percentage point.

7
It’s clear that the organizations making these moves in 2020—prioritizing AI/ML initiatives, increasing
spending, and expanding data science teams—expect their strategies will begin producing results sooner
rather than later. 2021 is shaping up to be extremely competitive in this regard, as the organizations
making serious investments in AI/ML will quickly outpace those still sitting on the sidelines. And these
forward-thinking organizations are putting particular emphasis on projects that help them meet the
expectations of the modern customer while still keeping their costs under control.

Trend 2: Organizations are expanding into a wider range of AI/ML use cases,
with particular focus on process automation and customer experience
Another trend that surfaced in our summer survey and became more pronounced in our 2021 survey
data is that organizations are focusing on AI/ML use cases that will reduce costs while improving the
customer experience. When respondents were asked about the different ways they’re applying AI/ML in
their organizations, customer experience and process automation rose to the top as some of the most
common use cases respondents selected. We also saw a dramatic (74%) year-on-year increase in
organizations that selected more than five use cases from the list of options in the survey.

Customer experience and process automation represent the top

AI/ML use cases

50% 40%
Generating Managing
customer insights
and intelligence 46 % inventory
Detecting
fraud
57% 43% 32%
44%
Improving customer Generating
Supply chain
experience financial
optimization
Increasing insights
long-term
customer 39%
33%
engagement Reducing
costs

48% Financial
planning
Interacting

45%
with customers

31% 49% Managing 1%

Automating
processes logistics

40% 22% 34%

Increasing Acquiring new

33%
customer customers
loyalty 27% 31%
Back office
automation

Recommender systems
Retaining customers Reducing customer churn Building brand awareness Other

Customer experience, engagement, loyalty Detecting fraud Finance Other

Automating processes and reducing costs Supply chain and logistics Customer acquisition

8
The percentage of respondents who indicated they have more than five
AI/ML use cases has increased 74% year-on-year

0%
52% 12%
No use cases

1-5 use cases No use cases 20%

>10 use cases
6%
>10 use cases

38%
2020 1-5 use cases 2021

29% 42%
6-10 use cases 6-10 use cases

Respondents were asked to select their use cases for AI/ML from a list of options, from which we calculated the total number of
indicated use cases per respondent. Respondents were given slightly different options for use cases between 2020 and 2021. See
methodology section for a list of use case options per year. The year-on-year percent difference was calculated with the underlying
data before being rounded to the nearest percentage point. Categories do not add up to 100% because they have been rounded to
the nearest percentage point.

What’s more, most organizations are overwhelmingly increasing usage for all of these applications of AI/ML,
but especially for customer experience improvement and process automation.

“
It’s illuminating to see what people are doing today with machine learning and the
impact it’s having on their organizations. When applied to automating processes, it’s saving
companies money by allowing them to execute at a larger scale with cheaper costs. Machine
learning is also having an impact at the top line, focusing on serving existing customers
better and driving customer acquisition. That’s where you get the real return on investment.
And that’s why we targeted Algorithmia where we did, to help these organizations unlock
value by moving models to production faster, at scale and cost efficiently.”

Kenny Daniel
Founder and CTO, Algorithmia

9
For nearly all use cases, 50% or more of organizations are increasing their
usage of AI/ML
0% 25% 50% 75% 100%

Improving customer experience

Automating processes
Generating customer insights and intelligence
Increasing long-term customer engagement
Generating financial insights
Acquiring new customers
Interacting with customers
Increasing customer loyalty
Managing logistics
Retaining customers
Supply chain optimization
Building brand awareness
Detecting fraud
Reducing costs
Back office automation
Financial planning
Recommender systems
Managing inventory
Reducing customer churn

0% 25% 50% 75% 100%

Increasing No change Decreasing

Respondents were asked to select how usage is changing for different use cases of AI/ML. Per each use case, this chart only
includes respondents who have that particular use case.

What does it all mean? The upheaval of 2020 has forced companies to be laser-focused on their most
important priorities, including AI/ML initiatives. According to the VUCA framework, we’re simultaneously
experiencing volatility, uncertainty, complexity, and ambiguity. This underlines the urgency of pursuing
AI/ML use cases that have a clear business ROI. The luxury of doing an AI/ML project for the sake of it
no longer exists.

This is why we’re seeing such emphasis on use cases that focus on customer experience and process
automation—they both have direct lines to tangible ROI. AI/ML use cases that enhance customer experience
drive top-line growth by capturing new customers and retaining existing customers. Process automation
generates short-term savings that boost the bottom line.

As organizations achieve tangible ROI wins in areas like these, they also build momentum for additional
innovation. We believe this is a key reason why we’re seeing organizations expand into a wider range of
use cases. These organizations are clearly moving past initial experimentation and scaling up their efforts
to achieve significant ROI.

10
Trend 3: Most organizations have more than 25 models in production, but there’s a
gap between AI/ML “haves” and “have-nots”
First, some good news: The average organization has already deployed a significant number of ML models to
production. More than half of all respondents have more than 25 models in production, and 40% have more
than 50.

However, when we compare the largest enterprises in our sample (organizations with more than
25K employees) with everyone else, we see a jarring divide. Only 7% of organizations with 25K or fewer
employees have more than 100 models in production, but among the very largest enterprises, 41% have
more than 100 models.

The world’s largest enterprises are dominating the high end of model scale
Organizations with 25K or fewer employees Organizations with more than 25K employees

50%

42% 41%
40 %

30%

24%
21%
20 %

17%
14% 13%
10% 11 %

10%
7%

0%

0-25 26-50 51-75 76-100 >100 0-25 26-50 51-75 76-100 >100

Models in production

11
 “
The tsunami of data generated by digital business can no longer be addressed with
human scale solutions. Organizations must move faster, and adopt new, ML and cloud-scale
services to deliver the real-time continuous intelligence they need to build and secure great
customer experiences. Don’t wait, start applying these new technologies today...
experiment, learn, improve and move forward.”

Dione Hedgepeth
Chief Customer Officer, Sumo Logic

As a previous McKinsey report found, the world’s largest technology

There’s a massive firms have benefited from both vast resources and early adoption.
opportunity to capitalize These companies have collectively spent billions of dollars on R&D and
on the next wave of AI/ other investment areas, and have been doing so for years. That gives
ML adoption in 2021. them a significant head-start.

That’s not to say that these companies are the only ones with the
right approach. Rather, it illustrates the importance of adopting AI/ML
as early as possible and making meaningful investments in its
long-term success.

As McKinsey wrote in its report, “evidence suggests that AI can

deliver real value to serious adopters and can be a powerful force for
disruption.” Moreover, “early AI adopters that combine strong digital
capability with proactive strategies have higher profit margins and
expect the performance gap with other firms to widen in the future.”

Some more good news: The barrier to entry is much lower today
than it was when the earliest adopters of AI/ML began investing in the
technology. We’re on the cusp of a second wave of adoption that is far
more accessible to organizations of all sizes and in all industries. AI/ML
is no longer the sole purview of technology titans with vast resources.

But a new gap may now be forming, and not between the largest
early adopters and everyone else. Rather, the next divide is occurring
(and growing) between enterprises that are making tangible AI/ML
investments now and those that continue to take little or no action.

There’s a massive opportunity to capitalize on the new next wave of

adoption, but not everyone is paying attention. Given that a growing
number of organizations of all types are prioritizing AI/ML—with
increasing budgets and staff to match—the message is clear:
The time to act is now.

12
Report theme 2: Organizations continue to face challenges across
the ML lifecycle, and struggle with cross-functional alignment
When organizations struggle to adopt and scale AI/ML, the reasons why have become increasingly clear.
Success across the entire ML lifecycle requires a high level of machine learning operations (MLOps) maturity.
We’ve outlined the path to that maturity in a whitepaper on the topic.

“
Algorithmia’s 2021 report clearly demonstrates the rising importance of AI for
companies across industries. Google was an early mover in this space and solved many
operational challenges associated with deploying ML models at scale. These challenges
still exist, and all organizations looking to adopt AI will have to solve them. Today, however,
companies can access pre-built solutions without having to grow this expertise in-house.
Organizations that can cleverly take advantage of these solutions and leverage AI will find
competitive advantages.

Anna Patterson
Founder and Managing Partner,
Gradient Ventures at Google

Trend 4: Governance is by far the top challenge for AI/ML deployment, with more than
half of all organizations ranking it as a concern
When we asked survey respondents to indicate the challenges they face when deploying models, we saw a
wide range of difficulties that occur across the ML lifecycle. However, the most common challenge by far was
with AI/ML governance, an issue that is primarily a concern at the end of the ML lifecycle when models have
already been developed and organizations need to minimize their risk.

13
“
The opportunity for AI is to allow banks to
provide services in much more personalized,
highly scalable, and customized ways. The
challenges include the ability to explain your
AI and to ensure confidentiality, since a lot
of the data in finance is personal information
or highly confidential. User trust is key. The
person on the other side wants to trust you
with their most valuable assets, and with
their most valuable information. The challenge
for MLOps is making governance and security
highly scalable in a time when investment in
AI and the number of production models we
are seeing is accelerating.”

Apoorv Saxena
Global Head of AI Technology, JP Morgan Chase & Co
56% of organizations struggle with governance, security, and
auditability issues
0% 20% 40% 60%

IT governance, security, auditability requirements 56%

Challenges with integration/compatibility of ML

technologies, programming languages, frameworks 49%

Monitoring model performance 37%

Frequent updates required to maintain

model quality, performance 36%

Duplication of effort/disconnection between teams

across the organization (e.g. data science and IT) 36%

Managing, allocating ML-related infrastructure costs 35%

Getting organizational alignment, senior buy-in 30%

Versioning, reproducibility in models 27%

Other 1%

0% 20% 40% 60%

67% of all organizations must comply with multiple regulations

0% 10% 20% 30% 40% 50%

ISO 27001 48%

OCC 2011-12,
SR 11-7 40%
HIPAA 36%
FISMA 34% 8%
No regulations
PCI 32%
SOC 32% 26%
One regulation

No regulations 8%
Other <1%
67%
Multiple regulations
0 %
10 %
20 %
30 %

Respondents were asked to identify regulations that their organizations are required to comply with. This chart shows the
percentage of respondents who selected each regulation. Respondents who didn’t select any regulations are shown in the “None”
category. For respondents who selected “Other”, “Other” was counted as a regulation. In the donut chart, categories do not add up
to 100% because they have been rounded to the nearest percentage point.

15
Respondents also indicated that they face a wide range of regulatory requirements, from SOC to
HIPAA to PCI. Nearly every organization is dealing with some level of compliance burden; when asked which
regulations they needed to comply with, 67% of respondents selected multiple regulations and only 8%
didn’t select a single regulation.

Since governance is a broad, longstanding term in IT, it’s important to give it clear meaning in this specific
context. We define AI/ML governance as the overall process for how an organization controls access,
implements policy, and tracks activity for models. From a regulatory compliance standpoint, governance
is a must-have to minimize organizational risk in the event of an audit.

But while regulatory compliance is important, it’s only one part of the governance equation. Effective
governance is the bedrock for minimizing risk to both an organization’s bottom line and to its brand.
Organizations with effective AI/ML governance not only have a fine-grained level of control and visibility into
how models operate in production, but they unlock operational efficiencies by integrating AI/ML governance
policies with the rest of their IT policies. They can document and version models, tracking both the inputs
and outputs of those models to understand all the variables that might affect model results. This enables
organizations to quickly identify and mitigate issues such as model drift that degrade the accuracy of results
and the performance of applications—issues that can directly impact the business’ bottom line and erode
customer trust in the brand over time.

Many organizations (more than half of those surveyed) are struggling with this important aspect of AI/ML
strategy. It’s also reasonable to assume that some organizations don’t yet realize they’re struggling with it;
governance is a late-lifecycle concern, so it doesn’t always become obvious to organizations until they’re in
the later stages of the ML lifecycle. As organizations head into 2021, they should consider whether they
need to focus more on AI/ML governance, even if it’s not currently a top concern.

“
Data scientists have traditionally been forced to enlist developers to help create custom
solutions for operationalizing and monitoring ML model performance. As the number of
models grows, so do the number of people involved and the complexity of operating and
maintaining these custom solutions with cross-functional teams. This forces a reexamination
of the entire approach. To address this challenge, organizations should look to products
that integrate operational tooling with off-the-shelf monitoring infrastructure so that data
scientists rapidly assemble ML monitoring solutions on their own. Simplifying the monitoring
process has the potential to eliminate model performance as a widespread challenge.”

Tim Hall
VP Products, InfluxData

16
Trend 5: The second greatest AI/ML challenge is technology integration and
compatibility, with 49% of organizations ranking it as a concern
While governance topped the list of AI/ML issues, there was the clear runner-up: 49% of respondents
said they experience challenges with the integration or compatibility of their ML technologies, programming
languages, or frameworks. This means organizations are still stumbling at the beginning of the AI/ML
lifecycle.

It’s not without reason: ML technology is unique and always developing. It’s not always ready for plug-and-
play integrations with other systems. Specialized hardware, such as GPUs, is critical to modern machine
learning, but presents a number of compatibility and software challenges.

The bottom line? This is still a relatively young industry and the technology is rapidly evolving. Staying current
with the AI/ML technology landscape requires regular monitoring and updates. That creates a burden on
organizations and their data science teams, especially those that take a do-it-yourself approach.

“
I think this survey shines a light on what I would call the ‘unsexy part of MLOps’ in that, more
teams cited major blockers to deploying models being security, governance, and audibility.”

Demetrios Brinkmann
Community Coordinator MLOps Community

Trend 6: Successful AI/ML initiatives require organizational alignment across

multiple decision-makers and business functions
Indeed, as the stakes and complexity around AI/ML initiatives grow, so does the need for collaborative,
cross-functional AI/ML teams that span departments. Strong, sustained results absolutely depend on this
team-building approach.

We asked survey participants to indicate who sets the priorities for AI/ML in their organizations, and got
widely diverse responses—though infrastructure and operations (I&O) leaders clearly rose to the top, with
CTOs, CIOs, and heads of data science following closely behind.

17
Successful AI/ML initiatives involve decision-makers from across the
organization
0% 20% 40% 60%

IT infrastructure and operations leader 57%

CTO / Head of Innovation 39%

CIO 35%
Head of Data Science / Head of Machine Learning 31%
Senior leadership (non-IT) 30%
Business intelligence and analytics 24%
Chief Data Officer 22%
Application/product owner 19%
DevOps 18%
Chief Risk Officer 12%
Other IT leader 11%

0% 20% 40% 60%

“
I have always found that when C-level executives are behind a technology initiative then it is
more likely to succeed. Any deep technology initiative is likely to have change rippling across
the organization, and C-level executives can enable the change management. So interesting
to see CTO and CIO in second and third place: I believe those organizations are more likely
to succeed with AI than others.”

Michael Azoff
Chief Analyst, Kisaco Research

18
One of the key takeaways that we see here is that AI/ML is no longer the sole or primary purview of data
scientists. Rather, a much broader cross-section of roles are leading or becoming involved with AI/ML
strategy and operations. Consider the different domains and responsibilities of these roles:

Infrastructure and operations

I&O is a well-established role in IT. As the name suggests, I&O is essentially the backbone that
makes fast, secure, and scalable systems possible. I&O pros typically take the lead on integrating
new technologies and initiatives into existing systems and processes in an organization, such as
the SDLC or CI/CD pipeline.

This means that I&O leaders are also increasingly responsible for scaling AI/ML initiatives in production.
They need to be convinced that AI/ML initiatives don’t introduce new risks or operational burdens. We
believe governance is becoming an increasing issue in AI/ML partially because I&O leaders (rightly)
take it seriously and are playing higher-profile roles in their organizations’ AI/ML strategies.

CIO

This is the top technical leadership role at most companies; this person and their direct reports are often
responsible for driving digital transformation within the company. CIOs are also increasingly strategic and
collaborative, meaning they’re working closely with their C-suite counterparts on the organization’s most
pressing goals. They’re also focused on AI/ML—and how to measure its success—from a leadership level.

Head of data science

This is an emerging role with a fluid definition, but the head of data science is generally responsible for
unlocking the value in an organization’s data for the business. The head of data science is responsible
for leading the team that owns the organization’s AI/ML capabilities and initiatives, and is responsible
for actually developing them and moving them forward. And the value proposition of this role is just
beginning to come to fruition, as cross-discipline teams work together to pursue quantifiable business
value from their data.

We believe one reason why tool and technology integration is a common concern is because this issue
falls within the purview of data science teams, who continue to be key players in AI/ML decision-making.

Business team

Business teams are those who create the products that actually use machine learning on the back end.
These professionals are responsible for driving market growth and helping the organization maintain
competitive edge, so they are commonly focused on the customer-centric use cases mentioned previously.

19
As you can see, there are now a wide range of roles and even organizational levels—from the C suite on
down—involved in AI/ML strategy and implementation, though I&O has become the major leader in setting
priorities.

The cross-functionality of AI/ML is also evident in the metrics organizations are using for success. We asked
organizations to indicate which metrics they use to evaluate the success of their AI/ML initiatives, and got a
wide range of responses.

Organizations are using a variety of success metrics for AI/ML initiatives

0% 25% 50% 75%

ML and statistical metrics

(e.g. f1 score, accuracy, precision) 60%
Time savings to the organization 55%
Cost savings to the organization 53%
Measures of additional value delivered
(e.g. product adoption, customer retention) 52%
ROI 40%
Measurement of bias and fairness 35%
Other 1%

0% 25% 50% 75%

“
In our conversations with F1000 CIOs and their teams, we find that they explore
ML technologies and ML use cases to gain a competitive advantage, and justify their
investments in the current term by communicating the efficiencies gained and the abilities to
drive better customer intimacy. Ultimately, as they crack the code and understand how best
to capitalize on their efforts, they begin to shift the justification of their investments through
business outcomes. [This report] is an outstanding visualization of the shift and recognizes
that organizations who lean forward, and use disruptive technologies to thrive through
disruptive times, will outlast organizations that remain risk- and change-averse.

Todd Tucker
VP, Standards and Education, Technology Business Management Council

20
Trend 7: Organizational alignment is the biggest gap in achieving AI/ML maturity
Despite the cross-functionality of AI/ML, when we asked organizations to rank their current level of
maturity against the maturity framework we outline in our whitepaper, organizational alignment was
the weakest point. While in most areas, a greater number of organizations ranked themselves as high
maturity than medium maturity, this trend was reversed when it comes to organizational alignment.

Organizational alignment is the biggest gap in achieving AI/ML maturity

60% 52%
51% 48%
44% 48
%
44
%
40 %
40 %
40%

20%
9% 8% 8% 8%
0%

Organizational Model development, Deployment, management, Governance and

alignment data, training operations compliance

Maturity level Low Medium High

It’s interesting that this was the lowest-ranked area of maturity in respondents’ self-assessments, because
in fact it’s the most foundational area for success. Not only does maturity in this area enable maturity in more
complex and technical areas (such as governance), but it is simply crucial to the success of any project due to
the cross-functional nature of AI/ML discussed above. Organizations that ignore this critical need will diminish
their potential ROI.

21
“
I believe that the competitive advantage of AI
exists at the intersection of data, technology,
and people.

You can only get value from your AI when it’s

woven with deep business accountability and
expertise. In order for enterprises to unlock the
real power of AI and realize its value, they must
first build a strong partnership between
data science and the business.”

Nir Kaldero
Chief Data, Analytics, and AI, NEORIS
Adjunct Executive, Global Data Science, CEMEX
Report theme 3: Despite increased budgets and hiring, organizations
are spending more time and resources—not less—on model deployment
Clearly, the economic disruption of the pandemic has brought a renewed sense of importance to AI/ML
initiatives in many organizations. As explained previously, organizations are moving AI/ML up their priority
lists and increasing budgets and staffing, too.

However, this doesn’t mean organizations can solve all the challenges they encounter in their AI/ML work
overnight. In fact, our survey revealed that as organizations increase their AI/ML investments, they are
actually spending more time and resources on model deployment, not less. We believe that organizations are
using their increased headcounts to manually scale AI/ML efforts rather than addressing underlying issues
with operational efficiency. But a growing AI/ML staff will have a much bigger impact if it’s able to focus on
data science instead of constantly paying down operational overhead.

Trend 8: The time required to deploy a model is increasing, with 64% of all
organizations taking a month or longer
Our data shows that the total timeline required to develop and deploy a model is significant. First off, we
polled respondents about the average amount of time it takes them to develop a trained model once they’ve
actually defined a use case (a new question in this year’s survey). A majority (66%) reported that this process
takes them a month or longer to complete.

Then, once a trained model has been developed, it must still be put into scaled production. When we asked
respondents about the average amount of time it takes them to do that, only 11% indicated they can deploy a
trained model to production within a week. The majority of organizations (64%) are taking a month or longer.

Only 11% of organizations can put a model into production within a week,
and 64% take a month or longer
0% 10% 20% 30%

1 day or less 1%
1 day–1 week 10%
1 week–1 month 25%
1 month–1 quarter 24%
1–2 quarters 18%
2–3 quarters 12%
3 quarters–1 year 6%
More than 1 year 3%
0% 25% 50% 75%

The total percentage of respondents who selected a month or longer was calculated with the underlying data before being
rounded to the nearest percentage point. Categories do not add up to 100% because they have been rounded to the nearest
percentage point.

23
This means that once organizations define a use case for AI/ML, they still face a significant timeline—months
to even years—to get to the point where they have a trained model developed for it and scaled to production.
What’s more, the time required to deploy a model (once it’s been developed) is actually getting longer. Since
we asked respondents this question in last year’s survey too, we were able to compare the results,
and found that it is indeed increasing year-on-year.

The time required to deploy a model is increasing year-on-year

2020 2021

50%

40%
40%
37%

30% 29%
25% 24%
20%
20%

11%
10 %

5% 5%
3%

0%

<1 week 1 wk. 1 mo. 1 qtr. >1 year <1 week 1 wk. 1 mo. 1 qtr. >1 year
to 1 mo. to 1 qtr. to 1 yr. to 1 mo. to 1 qtr. to 1 yr.
Time to deploy a model

Data from 2020 doesn’t include respondents who selected “I do not know or I am unsure”. Respondents were given slightly
different options for time ranges between 2020 and 2021. See the methodology section for a list of time range options per year,
and how they were aggregated for accurate year-on-year comparison. Categories do not add up to 100% because they have been
rounded to the nearest percentage point.

24
Trend 9: 38% of organizations spend more than 50% of their data scientists’ time
on deployment—and that only gets worse with scale
Our 2021 data also shows clearly that data scientists are spending too much of their valuable time on model
deployment. We asked respondents how much of their data scientists’ time is being spent on deploying
models, which we defined in the question as “prepping trained models and deploying them where they can
be consumed by apps or used with other models”. Our data showed that a full 38% of organizations are
spending more than 50% of their data scientists’ time on these tasks.

38% of organizations spend more than 50% of their data scientists’ time
on deployment
2020 2021

50 %

45%

40%
36% 37%

30%
27%

21%
20 %

17%

11%
10%
7%

0% 1%
0 %

0% 1-25% 26-50% 51-75% >75% 0% 1-25% 26-50% 51-75% >75%

Percentage of data scientists' time spent on deployment

Data from 2020 doesn’t include respondents who selected “I do not know or I am unsure” or “My team does not/has not deployed
models”, or who didn’t answer the question. Categories do not add up to 100% because they have been rounded to the nearest
percentage point.

The amount of time that data scientists spend on deployment actually increases with the number of models
in production too. Our data shows that organizations with the most models in production spend the greatest
percentage of their data scientists’ time on model deployment.

25
Organizations with more models spend more of their data scientists’ time
on deployment, not less
0-50 models in production 51-100 models in production >100 models in production
60%

40%

20%

0%
% 0% 5% 5% % 0% 5% 5% % 0% 5% 5%
0% 1-25 26-5 51-7 <7 0% 1-25 26-5 51-7 <7 0% 1-25 26-5 51-7 <7

Percentage of data scientists' time spent on deployment

Once again, this implies that organizations are using increased headcount to manually scale AI/ML.
Organizations that have scaled their models in this way are significantly limiting their potential ROI and
the long-term sustainability of their initiatives. They would be far better served by improving operational
efficiency and scale so their data scientists can focus on building innovative models—not performing
manual operational tasks.

“
Data scientists are expensive and increasingly hard to find, hire, and retain. Yet companies
are letting an increasingly large portion of that scarce capacity go to model deployment,
maintenance, and management. Data scientists are neither good at nor do they like doing
this stuff. This is just crazy. The companies that get this right will leave the ones that
continue on this course in their dust.”

H. P. Bunaes
Founder AI Powered Banking

26
Report theme 4: With increasingly complex infrastructure
needs, organizations report improved outcomes with third-party
MLOps solutions
As the AI/ML market matures, we see interesting trends in how organizations are approaching infrastructure.
First off, organizations have increasingly complex environments for deploying models. In our 2021 survey,
71% of all respondents indicated that they use a hybrid environment (consisting of more than one cloud
or on-premises infrastructure provider) to deploy models, and 42% of all respondents have a hybrid
environment consisting of both cloud and on-premises solutions.

71% of all organizations have hybrid environments, and 42% have a

combination of cloud and on-premises infrastructure

<1%
On-premises hybrid 3%
Single vendor

71 29 42
% % % on-premises

Hybrid Non-hybrid Hybrid cloud and

on-premises
25%
Single cloud

29%
Multi-cloud

Excludes respondents who selected “Other” (less than 1% of respondents). Categories do not add up to 100% because they have
been rounded to the nearest percentage point.

27
We also know that hybrid environments that include both cloud and on-premises solutions are becoming
more common. After excluding respondents who selected “I do not know or I am unsure” or “Other”,
16% of respondents in the 2020 survey indicated they have a hybrid environment consisting of both cloud
and on-premises solutions. This is significantly lower than the 42% of respondents who indicated this in
the 2021 survey.

And with increasingly complex environments come increasingly complex infrastructure needs. So how are
organizations handling those infrastructure needs? We asked respondents how their organizations approach
model deployment and management infrastructure, with four * different options:

1. We build and maintain our own system from scratch

2. We integrate open-source components into a system that is maintained in-house

3. We integrate commercial point solutions into a system that is maintained in-house

4. We use a third-party platform supported by a vendor

“
We see a recurring theme around tooling consolidation: two out of three teams we talk to,
say they are building an internal ML platform, which is really their attempt to standardize
tooling and process around ML development and deployment. Their effort is typically
motivated by a goal to cut model operating costs and the need for transparency over
production ML applications.”

Alessya Visnjic
CEO Why Labs
ex-Amazon Machine Learning

In the early days of AI/ML, any organization that wanted to deploy models at scale was essentially required
to build and maintain their own system from scratch. We see many organizations still gravitating toward this
approach, but this group is disproportionately skewed towards organizations with a large number of models.
Among organizations with more than 100 models, 60% chose to build and maintain their own systems from
scratch, but only 35% made this choice among other organizations. Our hypothesis is that these organizations
represent early adopters in the space, who have gravitated toward building their own solutions because it may
have been their only choice.

In contrast to the organizations that are building and maintaining their own systems from scratch are those
that are using a third-party solution—either integrating commercial point solutions into their systems or using
a third-party platform supported by a vendor. For the rest of this report, we will be comparing respondents
who selected option #1 above (which we’ll call “build from scratch”) with respondents who selected either
option #3 or option #4 (which we’ll call “buy a third-party solution”). This corresponds to the
build vs. buy decision for MLOps, which we’ve written about before.

28
*
We also provided a fifth option for “Other,” but it received less than 1% of all responses.
The market has matured significantly since the early-adopter enterprises first began investing in AI/ML. As
we’ve seen, organizations now have increasingly complex infrastructure environments, including hybrid cloud
and on-premises environments. Regardless of the specifics, IT infrastructure and applications tend to be far
more distributed than they were in the past. A third-party MLOps platform can be better equipped to handle
these complex environments, and shave off some of the total infrastructure costs as your models scale. We
see this reflected in our 2021 survey, as organizations that buy a third-party solution tend to spend less
money on infrastructure and less time on model deployment than organizations that build from scratch.

Trend 10: Organizations that use a third-party MLOps solution save money and
spend less time on model deployment than those that build their own solution
Our 2021 data found that buying a third-party solution yields better outcomes in multiple areas than
building from scratch. First off, we asked respondents to report their estimated annual infrastructure costs
for all models in production. Based on these responses, we created both a low estimate and a high estimate
for infrastructure costs an organization might expect. For both the low and high estimates, we saw cost
savings for organizations that buy a third-party solution. Indeed, according to our data, organizations that
buy a third-party solution spend an average of 19-21% less on infrastructure costs annually—and we have
reason to believe the actual cost savings seen in production could be much higher. This is because we used
a conservative method to calculate the average infrastructure cost*, which discounted the higher end of the
cost spectrum. Since a greater portion of organizations at that high end of the cost spectrum are building
their own solutions from scratch, their average costs were likely underestimated.

Buying a third-party solution costs 19-21% less than building your own
$6M
Respondents were asked to indicate their
$
5.5M -19% average annual infrastructure costs based on
difference
predefined ranges, such as “$51-$100K”. The
total average annual infrastructure cost was
then estimated as a range. The low estimate is
$5M based on the lower bound for each predefined
range (for example, $51K for “$51-$100K”).
$
4.5M The high estimate is based on the upper bound
for each predefined range (for example, $100K
-21% for “$51-$100K”). For the pre-defined range
$4M
$
3.9M difference
that represented the greatest cost (“more than
$10M”), the lower bound of the range was used
for both the high and low estimate. The percent
difference was calculated with the underlying
$
3.1M data before rounding to the nearest
$3M percentage point.

$2M
Low High Low High
estimate estimate estimate estimate
Build from scratch Buy a third-party solution

29
*
In both the high and low estimates, all organizations reporting more than $10M in annual infrastructure spend
were assumed to only be spending $10.000001M.
Organizations that buy a third-party solution also tend to spend less time on model deployment. When
we calculated the average percentage of data scientist time spent on model deployment, it was lower for
organizations that buy a third-party solution rather than building from scratch.

Organizations that buy a third-party solution spend less of their

data scientists’ time on model deployment
Build from scratch Buy a third-party solution

50 %
50%
46 %

40% 39% 40% 39%

32%
30%
30% 30%

23%
21%
20 %
20%
16 %

12%
10%
10% 10%

1% 1%
0 %
0%
0% 1-25% 26-50% 51-75% >75% 0% 1-25% 26-50% 51-75% >75%
Average % of data scientists’
Average % of data scientists’ time spent on model deployment time spent on model deployment

Respondents were asked to indicate the average percentage of their data scientists’ time spent on model deployment based on
predefined ranges, such as “26-50%”. The total average percentage of data scientist time spent on model deployment was then
estimated by using the lower bound of the data in each range (for example, 26% for “26-50%”) and averaging across respondents.
Categories do not add up to 100% because they have been rounded to the nearest percentage point.

30
Lastly, these organizations tend to get their models into production more rapidly. On average, the number
of days it takes them to put a trained model into scaled production is 31% lower than for organizations that
build from scratch.

The time required to deploy a model is 31% lower for organizations that
buy a third-party solution
Build from scratch Buy a third-party solution

30 %
29% 100 d
28% 92 d
90 d

25%
23% -31%
22% difference

20%
20% 64 d
60 d

15%
13%
12% 12%
11%
10% 9%
30 d
7%
6 %

5% 4%
2% 2%
1%
0% 0d
y

rs

rs

rs

ar

rs

rs

rs

ar
qt

qt
da

da
qt

qt

qt

qt

qt

qt
-1

-1
-1

-1

Average days
ye

ye
-1

-1
1

1
2

4
1d

1d
1

1
1w

1w

to model deployment
1-

2-

3-

1-

2-

3-
1m

1m
>

>

Respondents were asked to indicate the average time it takes to deploy a model based on predefined ranges, such as “1 month-
1 quarter”. The total average time to deploy a model was then estimated by using the lower bound of the data in each range (for
example, 1 month for “1 month-1 quarter”) and averaging across respondents. For comparability between ranges, all data was
converted to days before averaging it, with 1 month being equal to 30 days and 1 quarter being equal to 90 days. The percent
difference was calculated with the underlying data before rounding to the nearest percentage point. Categories do not add up
to 100% because they have been rounded to the nearest percentage point.

As organizations face increasingly complex infrastructure needs, third-party MLOps solutions can help
organizations save on infrastructure costs while also helping to solve one of the most prominent problems
organizations currently face: Speeding up model deployment while reducing the operational burden on data
science teams. Third-party solutions can also lower the barrier to entry for AI/ML. Truly, this field is no longer
limited to enterprises that can invest in building and maintaining their own infrastructure entirely from scratch.

31
 “
Today MLOps doesn’t need any introduction—it’s central to the AI strategy of any
organization. Organizations are now moving away from custom open-source tooling to
more standardized platforms for their model management, which will make this space even
more competitive. Looking forward, what’s even more interesting is how organizations are
handling complex issues of ethics, explainability, and bias as they roll out more AI. These
issues have far-reaching impacts on the way models are built and governed, and should
be important considerations for any organization using AI.”

Sacin Porwal
General Manager – AI Solution Engineering, Wipro

Conclusion
2021 will be a crucial year for AI/ML initiatives. There are plenty of reasons for optimism, including a new
sense of urgency and importance about AI/ML within many organizations, plus growing investments in terms
of prioritization, spending, and staffing. And AI/ML is far more accessible than ever before. You no longer need
to build and maintain your own infrastructure from scratch just to get started with AI/ML. Organizations of all
sizes, industries, and infrastructure needs can now get started with AI/ML more quickly, or scale their
existing AI/ML efforts with greater ease.

What’s also clear is that the organizations that will reap the greatest benefits from AI/ML in 2021 are
those that invest in operational efficiency and scale. Those organizations will be able to more effectively
direct their AI/ML investments to the efforts that drive the most significant top- and bottom-line impacts
for their businesses.

2021 will certainly be a year when the gap grows between those organizations that take bold steps to scale
their AI/ML initiatives, and those that get mired in operational and organizational issues. It’s time to act—your
AI/ML future depends on it.

32
“
COVID-19 has caused rapid change which has challenged
our assumptions in many areas. In this rapidly changing
environment, organizations are rethinking their investments
and seeing the importance of AI/ML to drive revenue
and efficiency during uncertain times.

Before the pandemic, the top concern for organizations

pursuing AI/ML initiatives was a lack of skilled in-house
talent. Today, organizations are worrying more about how
to get ML models into production faster and how to ensure
their performance over time.

While we don’t want to marginalize these issues, I am

encouraged by the fact that the type of challenges have
more to do with how to maximize the value of AI/ML
investments as opposed to whether or not a company
can pursue them at all.”

Diego Oppenheimer
CEO and Founder, Algorithmia
Methodology
The purpose of Algorithmia’s 2021 enterprise trends in machine learning report is to report on the latest
developments and trends in enterprise machine learning and how they have evolved over the past year.
The report is based on data that Algorithmia collected in November 2020 in a survey effort that returned
403 responses.

The survey asked 29 questions about AI/ML initiatives, challenges, infrastructure, company demographics,
and more. The survey questions were developed by Algorithmia, and an independent third-party company
conducted the survey on Algorithmia’s behalf to ensure survey attribution anonymity and remove bias for
or against Algorithmia on the part of the respondents.

Respondents voluntarily participated in the survey in exchange for access to content or a service, such as
free Wi-Fi. Respondents received no monetary payment for their participation.

The third party screened participants using the following questions:

• What is your company size? (Only respondents at companies with $100M+ in revenue were included)

• Which best describes your role? (Respondents with roles of Consultant or Student were excluded)

• Are you involved with artificial intelligence (AI) and/or machine learning (ML) projects at your company?
(Only respondents who answered “Yes” were included)

In this way, Algorithmia amassed a group of 403 individuals with a level of insight into the machine learning
efforts of their companies across a random sampling of industries and machine learning maturity levels.

In all charts and analysis, percentages have been rounded to the nearest percentage point.

Changes from previous years

Algorithmia’s annual report about enterprise machine learning is an evolving project and we seek to make
improvements every year. We made multiple changes to this year’s survey and report to improve both the
reliability and relevance of our insights to AI/ML leaders.

This year, we limited the survey to respondents at companies with $100M or more in revenue. This resulted
in a slight shift towards organizations with more than 1,000 employees, with respondents being distributed
accordingly:

• Fewer than 500 employees: 3% (down from 18% in 2020)

• 500-999 employees: 17% (flat year-on-year)

• 1,000-4,999 employees: 33% (up from 29% in 2020)

• 5,000-9,999 employees: 19% (up from 14% in 2020)

• 10,000-25,000 employees: 11% (up from 9% in 2020)

• More than 25,000 employees: 18% (up from 13% in 2020)

We made this change to improve the relevance of results to enterprise IT environments, and plan to
continue this focus in subsequent years. As we establish a consistent baseline of company sizes in future
years, we expect the reliability of our results to further improve.

34
In previous years, we have also distributed the survey to individuals who have engaged with Algorithmia
through various channels in the past, such as by attending a company webinar, downloading a whitepaper, or
meeting with our team at an industry trade show. While we still sent a survey to this audience this year, we
ultimately did not include the results in the final report. To remove possible sources of bias, we only included
responses from the blind survey in this year’s report. When referencing data from last year’s report, we also
only used data from the blind survey conducted that year. However, while we did not include the data from
this version of the survey, we did use its results to analyze trends and further bolster our confidence in
our findings.

To gain deeper and more relevant insights from the survey, we also added and modified multiple questions in
the survey. Anytime this may have impacted year-on-year comparisons, we have included a footnote in the
relevant chart.

Specifically, in both our 2020 and 2021 surveys, respondents were asked to select the ways that AI/ML is
being used at their organizations. Respondents were given slightly different options between these two
years, as follows:

• Included in both years: Improving customer experience, Retaining customers, Reducing

customer churn, Interacting with customers, Increasing long-term customer engagement, Increasing
customer loyalty, Reducing costs, Acquiring new customers, Building brand awareness, Recommender
systems, Other

• Included in 2021 survey only: Automating processes, Back office automation, Managing inventory,
Managing logistics, Supply chain optimization, Financial planning, Generating financial insights

• Included in 2020 survey only: Increasing customer satisfaction, Processing automation for internal
organization, Increasing conversion rates, Predicting demand fluctuations, Filtering assets and content

In a question about the infrastructure used to deploy models, respondents were also given slightly different
options in 2020 and 2021, as follows:

• Included in both years: AWS, Azure, Google Cloud Platform, Other

• Included in 2021 survey only: VMWare-based, OpenShift-based

• Included in 2020 survey only: On-premises, A mix of cloud providers, A mix of cloud and
on-premises solutions, I do not know or I am unsure

35
Lastly, when asked about the time it takes to put a trained model into scaled production, respondents were
given slightly different options in 2020 and 2021. When comparing data year-on-year, data was aggregated
accordingly:

• 2021 options: 1 day or less (aggregated to “1 week or less”), 1 day-1 week (aggregated to “1 week
or less”), 1 week-1 month, 1 month-1 quarter, 1-2 quarters (aggregated to “1 quarter-1 year”), 2-3
quarters (aggregated to “1 quarter-1 year”), 3 quarters-1 year (aggregated to “1 quarter-1 year”),
More than 1 year

• 2020 options: 0-7 days (aggregated to “1 week or less”), 8-30 days (aggregated to “1 week-1 month”),
31-90 days (aggregated to “1 month-1 quarter”), 91-365 days (aggregated to “1 quarter-1 year”),
More than 1 year, I do not know or I am unsure

We will continue to conduct the annual survey to increase the breadth of our understanding of machine
learning in the enterprise, and share our insights into how the industry is evolving. As we continue to build
on this work in subsequent years, we aim to reach ever more relevant and applicable insights to help AI/ML
leaders drive innovation in this space.

36
About Algorithmia
Algorithmia is machine learning operations software that manages all stages of the ML lifecycle within
existing operational processes. Put models into production quickly, securely, and cost-effectively.

Unlike inefficient and expensive do-it-yourself solutions that lock users into specific technology stacks,
Algorithmia automates ML deployment, optimizes collaboration between operations and development,
leverages existing SDLC and CI/CD systems, and provides advanced security and governance.

Over 120,000 engineers and data scientists have used Algorithmia’s platform to date, including the
United Nations, government intelligence agencies, and Fortune 500 companies.

Learn how Algorithmia can help solve your most pressing ML challenges and put you in position for
scalable success. Visit algorithmia.com/product.

37
 About the cover
The cover image is a parallel set chart—similar to a Sankey diagram. Each line-set
represents a specific data category. The width of each line-set’s path is determined
by the proportional amount of the category total.

The line-set on the left depicts survey participants’ current state of maturity in the
following area: Organizational alignment for AI/ML initiatives. The line-set on the right
displays approximately how many ML models their organization has in production.

26-50 models

Medium maturity

11-25 models

51-75 models

High maturity
More than 100

6-10
Low maturity

76-100

1-5

None
Copyright © 2020 Algorithmia, Inc. All Rights Reserved.