A Product Manger’s Guide To Plan GTM of An Enterprise AI Product
“Telling cats from dogs is easy. It’s the what-ifs that get problematic”, says Susan Athey, the Economics of Technology Professor at Stanford School of Business in this article. Athey warns that decision makers in Enterprise, influenced by media hype on AI, can be led astray by misunderstanding what machine learning models can or cannot do.
A key piece of advice to keep in mind while you are building your GTM strategy for AI products targeted to Enterprises.
And yes, we do need a GTM strategy — AI will not sell itself, at least not yet.
In my last blog, I have outlined design and engineering facing product management responsibilities to build an Enterprise AI product. I want to focus on outbound facing Go-To-Market plans in this blog.
GTM strategy is a set of action plans that specify how a company will reach target customers and establish market position. Target market, buyer persona, value proposition, pricing strategy, marketing and promo, channel strategy — these are parts of GTM plan. And good product mangers do know this.
So, what is different when it comes to Enterprise AI products?
For Enterprise AI product, articulating value proposition needs a more nuanced understanding of buyer persona.
In the complex Enterprise taxonomy, few examples of emerging Enterprise AI products by function could be :
Employee Productivity | Business Intelligence | Customer Management | HR & Talent Management | B2B Sales & Marketing | Consumer Insights | Finance & Operations | Security & Risk | Engineering Development | Customer Experience Management
Although products built for different functions need customized GTM strategy, one common requirement is to be able to articulate AI-driven competitive advantage the product is bringing to the Enterprise.
Is it cost, growth, performance, brand/status, risk reduction, accessibility, convenience/usability, agility, customer delight? What value does AI deliver to your customer? What process transformation driven by AI will solve their problem?
Customer segmentation is one of the ways to answer this question. I like to characterize target enterprises based on this following framework.
Context — Where the target enterprise customer is in terms of overall AI strategy? Investigate how AI-enthusiastic this particular vertical is relative to others.
Challenge — What are the specific challenges AI will solve? Understand view points from all related parties, product users to budget approvers.
Capabilities — Where the target enterprise customer is in terms of AI capabilities? Construct a novel index to measure the target customer’s AI potential.
Consequences — What are the potential implications for this particular vertical to participate in AI ecosystem? Visualize a long-term trajectory of AI development and understand issues of safety and ethics, labor economics, and social governance.
Context :
One of the major challenges when talking about machine learning to Enterprise buyers is to find a middle ground between mathematical expressions Data scientists throw at them on one hand and the fantasies about general AI that media is talking about on the other. Find out which end of the belief system the target vertical or target customer belongs to.
Machine learning and artificial intelligence has shown transformative potential in the world of vision, speech and text. Image recognition started to achieve human-level accuracy at complex tasks. For example, Ever.ai automatically organizes 12 billion photos and videos for tens of millions of users in 95 countries and offers face detection, clustering, verification and identification APIs and mobile SDKs. There are plenty of examples of AI success in speech and text processing.
Enterprise world, on the other hand, is mostly non-vision today where language, context and reasoning play important roles. It is a much harder task to reverse-engineer the text back to ideas conveyed in sentences and paragraphs.
Machine learning will enable enterprises to get a better answer for a set of specific questions they all already have and based on the data that already exists. It will match resumes to job descriptions augmenting Applicant Tracking Systems. It will find patterns in malware augmenting rule-based Enterprise security breach detection. It will enhance customer experience by ranking user feedbacks from social media based on relevance.
Understanding target customer or vertical’s expectation of machine learning will help reduce GTM cycle.
Challenge :
According to a survey from MIT Sloan Management Review and BCG, “Almost 85% enterprise companies believe AI will allow their companies to obtain or sustain a competitive advantage. But only about one in five companies has incorporated AI in some offerings or processes.”
A major barrier is lack of understanding of what problems are worth solving. If your target customer is one of those organizations taking technology-first approach (“Let’s build a data-lake and find a project where we can use machine learning”) — a sure sign to stay away — Proof of Concept validation will be really long.
Once a right problem gets selected based on practical approach to solve a business problem (cutting down 6 months of knowledge worker’s time by building a specific AI-assisted decision-making process), make sure to articulate the value new product brings, focus on opportunity over savings — (decisions with higher accuracy and consistency, upside of not missing revenue due to 6 months delay in decision-making, wrong decisions causing interruption in knowledge worker’s time or damaging company’s brand reputation).
Finally, explainability of decision-making is crucial in enterprise applications. Why the algorithm rejected a potential candidate, approved loan to some undeserving candidates or flagged a transaction as fraudulent? Champion users of your AI product in target enterprise should be able to explain that.
Capabilities :
There are large gaps between today’s AI leaders and laggards across Industries, Nation States and Enterprises. Access to data, Digital maturity and C-suite sponsorship to build a sense of urgency in adopting AI towards core activities — are the main drivers that create the gap.
We keep hearing Googles, Amazons and FaceBooks of the world have all the data for ML/AI model building. What else could an Enterprise AI product offer without the data that those behemoths own?
Data is not fungible. Every industry has its unique data sets and regulatory requirements. One way to evaluate a target customer or industry or use case is to figure out — do they have relevant data? Have they developed a habit of collecting required data for training ML/AI models?
Frontier AI sectors are Finance, Automotive, Tech and Telco whereas the slower adopters are Healthcare, Construction, Travel, Education.
Having a model of AI potential index as a measure of an Enterprise’s capability to adopt your product is useful.
Consequences :
Commercial applications of AI/ML will transform businesses, destroy many jobs in their current forms and create new job categories. Not all industries or countries would be affected in a consistent manner. Automation has significantly transformed manufacturing, but barely touched construction ($12 Trillion construction industry ~13% of Global GDP) up until recently. Jobs in Spain are lot more vulnerable than Norway as of now, as Norway has already automated many tasks and processes across multiple industry sectors.
A new OECD analysis finds 14% of jobs across 32 countries are highly vulnerable (70% chance of automation) and a further 32% has a chance of being automated, with a probability between 50% and 70%. Low-skilled jobs such as, food preparation, cleaning, agricultural labour, personal services are facing higher risk of automation.
Whether you are building an AI product that cuts down 360,000 hours of manual review of contract documents to seconds (COiN platform from JPMorgan Chase), accelerates sales and customer flow with human-friendly AI assistant (tact.ai) or changes velocity of service offering by Telcos/IaaS/PaaS providers with intelligent AI-assisted operations, there are distinct winners (those who will gain from automation) and sufferers (those who have to re-invent and re-purpose themselves) in this process.
Being mindful about a broad range of social and economic consequences and building products and features that augment your target Enterprise workforce will definitely improve adoption.
