Why it’s so hard to market enterprise AI/ML products and what to do about it

In 2019, I led the sales team and growth strategy for a venture-backed AI company called atSpoke. The company, which Okta ultimately acquired, used AI to augment traditional IT services management and internal company communication.

At a very early stage, our conversion rate was high. As long as our sales team could talk to a prospect — and that prospect spent time with the product — they would more often than not become a customer. The problem was getting enough strong prospects to connect with the sales team.

The traditional SaaS playbook for demand generation didn’t work. Buying ads and building communities focused on “AI” were both expensive and drew in enthusiasts who lacked buying power. Buying search terms for our specific value propositions — e.g., “auto-routing requests” — didn’t work because the concepts were new and no one was searching for those terms. Finally, terms like “workflows” and “ticketing,” which were more common, brought us into direct competition with whales like ServiceNow and Zendesk.

In my role advising growth-stage enterprise tech companies as part of B Capital Group’s platform team, I observe similar dynamics across nearly every AI, ML and advanced predictive analytics companies I speak with. Healthy pipeline generation is the bugbear of this industry, yet there is very little content on how to address it.

Maintain a link to categories that are well known in early messaging, even if the category is not the core of your value proposition or why people will eventually sign a contract.

There are four key challenges that stand in the way of demand generation for AI and ML companies and tactics for addressing those challenges. While there is no silver bullet, no secret AI buyer conference in Santa Barbara or ML enthusiast Reddit thread, these tips should help you structure your approach to marketing.

Challenge 1: AI and ML categories are still being defined

If you’re reading this, you likely know the story of Salesforce and “SaaS” as a category, but the brilliance bears repeating. When the company started in 1999, software as a service didn’t exist. In the early days, no one was thinking, “I need to find a SaaS CRM solution.” The business press called the company an “online software service” or a “web service.”

Salesforce’s early marketing focused on the problems of traditional sales software. The company memorably staged an “end of software” protest in 2000. (Salesforce still uses that messaging.) CEO Marc Benioff also made a point of repeating the term “software as a service” until it caught on. Salesforce created the category they dominated.

AI and ML companies face a similar dynamic. While terms like machine learning are not new, specific solutions areas like “decision intelligence” don’t fall within a clear category. In fact, even grouping “AI/ML” companies is awkward, as there is so much crossover with business intelligence (BI), data, predictive analytics and automation. Companies in even newer categories can map to terms like continuous integration or container management.

Tactically, this means AI and ML solutions can’t rely on prospects typing specific solutions like “label quality” into, for example, Google.

To succeed, these companies need to stay in “category creation” mode. Practically, staying in this mode means being hyperfocused on the problem statement versus the solution in copy and assets.

For Hypersonix, a B Cap portfolio company, that might look like posing the question, “How much money are you losing by having a traditional analytics team determine your promotional strategy?” Buyers may not know the specific solution exists, but they certainly feel the pain.

Another tactic is language diligence. Be deliberate in the new terms you create and repeat those terms across your material. Examples of companies doing this well include Segment, with its self-created category of “Customer Development Platform;” Clari, with “Revenue Intelligence;” and — of course — Salesforce, with “SaaS.”

Finally, maintain a link to categories that are well known in early messaging, even if the category is not the core of your value proposition or why people will eventually sign a contract. For atSpoke, that was “filing tickets from Slack,” even though later sales pitches focused on automated answers and workflows.

Challenge 2: It’s a product, but it’s really a pipeline

AI or ML solutions are rarely — if ever — standalone. They are part of a pipeline of data tooling that moves from source to insight or action. For example, a company that intelligently forecasts revenue might actually provide data visualization, model training and application, but require alternative tooling for data preparation and transformation.

Compounding the issue is that no two data pipelines are exactly the same. Datasources can vary from Hadoop and Snowflake to marketing automation software; architectures can change from batch and streaming to Lambda; targets or outputs can be bespoke dashboards or common BI tools. This complexity and need for customization drives the high “services” revenue for AI and ML companies (but that is a separate topic).

Allen Chen, who leads product and engineering for BCG’s data science platform, summarizes these issues well:

In the software development lifecycle, the lines are extremely clear — developers move from Github to CircleCI, for example; in AI/ML those lines are not as defined. So then you end up having a conversation about the entire lifecycle around your tool — even if your product is only 10% of the full solution.

This is a problem for marketing. It is very hard to have a simple value proposition that resonates with customers or adheres to the tech marketing mantra, KISS (keep it simple, stupid). Ads risk being a mouthful of “ifs,” “thens” and clarifications; this is never a recipe for success, even with a technical product or target.

One way to address this is to avoid the complexity altogether. C3.ai, a publicly traded AI company, has billboards over the San Francisco area that read, “This Is Enterprise AI.” I like to think this is a nod to the complexity of the company’s offering. By keeping it so radically simple, C3.ai can increase brand recognition and drive interested people to marketing websites or content that more accurately reflect the breadth and depth of the company’s solutions.

It is worth mentioning that C3.ai likely faces the same pipeline challenge as smaller AI startups. Per their 2021 annual report, the company spent $53 million on sales and marketing expenses and had 40 new customers.

Another approach is to take the pipeline out of the early GTM process to focus on the solution. For example, Google makes Tensorflow tutorials interactive — you can just open up a notebook and try the APIs. For other AI/ML companies, similar strategies work, like creating sandbox environments that allow data scientists to simply drop python code into a Jupyter Notebook.

Challenge 3: The space is both new and crowded

There are more than 9,000 machine learning startups and companies, per Crunchbase. Firstmark Capital’s Machine Learning, Artificial Intelligence and Data (MAD) landscape provides a glimpse into the Cambrian explosion of companies in the space. To illustrate, the landscape outlines 19 ML platforms, 14 data science notebooks and 13 AI synthetic media companies — and these are all at scale.

Further, open source libraries like Facebook’s PyTorch and Google’s Tensorflow can directly compete with a lot of solutions. So AI and ML companies face competition from free tools that are already part of developers’ tech stacks (and backed by real heavies).

To address that, you need to get specific — terms like AI and ML have practically zero weight with buyers anymore. Choosing a highly specific vertical and problem statement and tailoring your messaging around that industry is table stakes. At Splunk, we had hundreds of industry specific campaigns and articles around, for example, logs for outages.

Finally, while your messaging needs to be industry-specific, don’t try to go head-to-head with the open source options in early marketing material. As Chen explains, “You need to have a value proposition that is higher up the stack; don’t look to pry [data scientists’] tools out of their hands … you will not convince a data scientist that your thing is better than PyTorch.”

Challenge 4: Many stakeholders, many priorities

The final headwind is that AI and ML companies need to market to multiple, very different stakeholders all at once. Unlike, say, a human resources information system that is primarily purchased by an HR department, an analytics solution will have to satisfy a business owner, a data science team and developers without including function-specific roles (like head of pricing) or procurement.

All these sets of buyers are growing savvier. Business owners will demand to see a tight connection between product and value. About 85% of AI projects failed to deliver impact, according to two recent Gartner reports. Analysts and engineers will want to see fast implementation timelines and credible promises of smooth integration.

To meet these varied demands, take a multipronged approach. Target business buyers with meaningful value propositions based on the problem statement.  For more technical stakeholders like developers, offer free tooling and documentation. DataRobot, for example, not only divides its marketing website by solution and industry (quite common across B2B SaaS) but also by roles such as “Business Analysts” and “Software Engineers.”

Final thoughts

Marketing in this field will likely follow a course similar to more established categories like DevOps, IT services management and CRM. In these spaces there are clearer categories to anchor to but far more competition for eyeballs and checkbooks. Buyers tend to be more educated. Tactically, this calls for niche marketing messaging, investment in branding and focus on optimization.

While the current environment is complex, in many ways, it can be freeing for your marketing strategy. Your company can play a role in defining the space it will one day win.

Disclaimer: B Capital, where Mike is employed, holds a financial stake in DataRobot, Hypersonix, Clari and Labelbox. Mike personally holds stock in Splunk.