This interview is with Kuber Sharma, Enterprise AI Strategist and Go-to-Market Leader.
Kuber, as an Enterprise AI Strategist and Go-to-Market Leader in enterprise software with experience at Microsoft, Salesforce, and Tableau, how do you describe your current focus and the specific enterprise problems you’re best at solving?
My focus right now is the gap between what AI can do and what enterprises actually deploy at scale. I spend most of my time with organizations that have been running automation for a few years, have real production use cases, but are hitting a wall when AI enters the picture.
The problems I know best:
- How do you take a process that a human performs with judgment and translate it into something an agentic system can handle reliably?
- How do you sell that internally, because enterprise AI is as much an organizational buy-in problem as a technical one?
My background across Microsoft, Salesforce, and Tableau gives me a useful vantage point. I’ve seen what happens when you ship capability before the market is ready to use it (Tableau’s early days with self-service BI), what happens when you nail positioning but underestimate change management (CRM rollouts at Salesforce), and what happens when you have a platform play but no clear entry point for buyers (Microsoft’s everything-is-AI era).
What I’m actually good at:
- Figuring out who in a customer organization will champion a new capability before the deal closes.
- Building GTM motions that survive contact with real enterprise procurement cycles.
- Translating what engineers built into something a CFO will approve.
I don’t love the term “AI strategist,” honestly. What I do is help companies stop piloting AI and start running it.
What pivotal choices or moments moved you from classic product and B2B marketing into leading AI strategy and agentic automation at enterprise scale?
Honestly, the move wasn’t a dramatic pivot. It was more like the category came to me.
When I was doing product marketing at Tableau, the big debate was whether business users could really self-serve with data or whether they’d always need an analyst in the room. Same pattern at Salesforce: would CRM actually change how reps work, or just become another system they had to log into. The answer in both cases was: only if you got the organizational change right, not just the software.
By the time I was deep in automation, I realized I’d been working on the same underlying problem my whole career. Capability doesn’t automatically change behavior. Someone has to figure out the wedge, the right use case, the internal champion, and the story that moves a deal from proof of concept to production.
The shift to agentic AI sharpened the stakes. Agentic systems make decisions. That changes the risk calculus for enterprise buyers in a way that earlier automation didn’t. So suddenly the go-to-market problem became a lot more interesting. You’re not just selling efficiency anymore; you’re selling a new kind of trust between humans and software.
I didn’t set out to be an AI strategy person. I just kept following the hardest version of the problem I already knew how to work on.
You’ve said the first sales meeting now happens inside a model; what is the single most high-leverage change you make to a product’s website to win that meeting?
The single change? Stop writing for SEO and start writing for extraction.
When a buyer asks a model to compare automation platforms, it’s pulling from whatever text it can cleanly summarize. Dense, jargon-heavy marketing copy that humans skim past gets ignored by models too. What survives is specific, direct language: this product does X for companies with Y characteristics, and it’s different from alternatives because of Z.
The highest-leverage thing you can do is rewrite your above-the-fold copy and your positioning page to make claims that are literally quotable. Not “AI-powered intelligent automation for the digital enterprise” but something a model can actually surface when a VP of Operations asks what the difference is between platform A and platform B.
Secondary to that:
- Structured FAQ pages that explicitly answer the comparison and objection questions buyers actually have. Not SEO FAQs—the real ones.
- How is this different from the competitor they’re already using?
- What does implementation actually cost?
- Who in their organization owns this after go-live?
Most websites still optimize for the human who’s already decided to visit. The harder problem now is being findable and quotable when the buyer hasn’t decided to visit yet, and a model is doing the first round of due diligence for them.
If I had one hour with a product marketing team today, I’d spend it rewriting the three sentences at the top of their website to be specific enough that an LLM could quote them accurately.
Staying with measurement, when you instrument AI-sourced demand, what one metric or diagnostic view tells you the strategy is working before revenue shows up?
The metric I track most is branded search velocity relative to paid spend. When AI starts sending buyers to you, the first thing they do is go verify youre real. If your branded search volume and direct traffic are ticking up without a corresponding increase in paid budget, something has changed in how youre being discovered. Thats usually AI referral, even if your attribution model cant label it.
Before revenue shows up, the honest leading indicator is demo-request quality not volume. Quality. Buyers who start their research in a model come in more informed. They arrive at their first call having already asked ChatGPT or Perplexity to explain the landscape, so they ask sharper questions earlier in the cycle. When my sales team tells me inbound prospects are coming in more prepared than usual, thats often a signal that were getting AI-sourced traffic even if it doesnt show up in our dashboards.
The attribution stack is the real problem. Most B2B tools were built for last-click and cookies. Theyre not designed to detect that a buyers journey started inside a model.
So I add something stupidly simple: a free-text “how did you first hear about us?” field on the demo form, and I actually read the responses. The signal is there. Its just not in your analytics platform yet.
The one metric that tells me the strategy is working before revenue materializes is direct-traffic growth plus demo-quality score, tracked in parallel against a 60-day lag.
Shifting from GTM to product, when you turn a legacy enterprise workflow into an agentic product, what is the first capability you ship to earn buyer trust, and why?
The first capability you ship is always audit and override—not the automation itself. The ability to see what the agent did and to undo it.
The barrier to deploying agentic systems in the enterprise is not technical capability. It is accountability. When a process goes wrong today, there is a human to point at. When an agent does something unexpected, the buyer needs to know they can catch it, explain it to their CFO, and fix it without calling the vendor.
So the first capability that actually earns trust is a clean activity log that anyone in the organization can read, plus a manual override that doesn’t require IT to execute. Not because those are the features buyers ask for in demos—they usually aren’t. But they are the features that determine whether legal, compliance, and the CIO sign off on moving from pilot to production.
What I’ve seen fail consistently are companies that ship the impressive automation first. The pilot works great. The demo is clean. Then the first edge case hits in production, no one can explain why the agent made the decision it did, and the whole program gets paused for a six-month internal review.
If you want to earn trust fast, let them watch before they let go. Show them every decision the agent made in plain language. Make the escape hatch visible and easy. Autonomy can come in version two. Transparency has to come in version one.
The teams that ship trust infrastructure first close faster and expand faster. It’s not a nice-to-have.
On the market-reading side, drawing on your analyst relations experience, what early signal from an analyst briefing or inquiry has most reliably predicted market pull for an AI product?
The most reliable signal I’ve seen is when an analyst stops asking you to justify the category and starts asking how you compare to specific competitors.
Early in a product’s life, analyst briefings are full of “why would an enterprise buy this?” and “what’s the actual use case here?” When those questions stop, it usually means the analyst has internalized the need independently. Someone else convinced them, or enough of their enterprise clients called in asking about it. Either way, the market is moving without you having to push it.
The second signal is when they start using your language in their inquiry questions to your competitors. Analysts run hundreds of briefings. If they’re using framing they first heard from you, it means the framing stuck. I’ve been in competitive situations where the analyst asked almost exactly the question I taught them to ask two quarters earlier. That’s when you know you’re winning the narrative, not just the briefing.
On the AI side specifically, the signal I watch for now is the shift from “is this real?” to “how do enterprises govern this?” When Gartner and Forrester start fielding governance and compliance questions about agentic AI from their clients, it means buyers have moved past skepticism and into procurement mode. They’re not asking whether to buy, they’re asking how to buy safely.
That shift from capability questions to governance questions is the clearest early signal I know of for real enterprise market pull.
Taking those signals into market entry, for a new vertical like insurance or HR tech, how do you pick the beachhead use case that will unlock the first ten enterprise logos?
I look for three things in a beachhead use case, and they’re not what most GTM teams look for.
-
First: the process has a single, clear owner. Not “the operations team.” A specific role: someone who will be personally accountable for the outcome and can champion the purchase without needing buy-in from four other departments. In insurance, that’s often the claims processing supervisor. In HR tech, it’s usually the recruiting ops lead.
-
Second: the ROI is countable, not estimated. I want a use case where you can say “this process takes X hours and costs Y dollars per unit” before you automate it, and then measure the delta after. If the benefit is “improved decision quality” or “better employee experience,” it will stall in procurement. Finance needs a number they can defend to their CFO.
-
Third: when it fails, it’s recoverable. The first ten logos in a new vertical aren’t just revenue; they’re reference stories and the foundation of your pricing model. You need use cases where an edge case or a bad outcome doesn’t make the news and doesn’t destroy the relationship. Claims adjudication in insurance, for example, is high-stakes enough that a mistake has real consequences. Start with something upstream: triage, document extraction, pre-screening.
The beachhead that unlocks the first ten logos is almost never the most exciting use case in the vertical. It’s the most defensible one—the one where you can guarantee a before-and-after comparison that holds up in a boardroom six months later.
Once the beachhead is defined, you’ve built on ecosystems like Salesforce and Azure; which integration or platform move has most accelerated adoption for you, and what made it succeed?
The moves that accelerated adoption most had one thing in common: they made the product show up inside something the buyer already opened every day, rather than asking them to open something new.
When Tableau was in the Salesforce AppExchange, it changed the sales motion. Instead of a separate evaluation cycle, data visualization became part of a broader CRM conversation. The ‘how do we get this past procurement’ question got a lot easier because Salesforce was already approved.
The same logic applies to Azure Marketplace. When your product is listed there, you’re not just another software vendor. You’re a line item that enterprise buyers can apply against committed cloud spend they’ve already budgeted. That changes who shows up to the evaluation and how fast they can say yes.
But here’s what actually made those integrations succeed:
- Someone on both sides had a real incentive to make it work.
- The Salesforce AE had a reason to recommend the add-on.
- The Microsoft seller got territory credit.
- Without that co-sell alignment, marketplace listings sit there and collect no traffic.
The lesson I’d take into any ecosystem play: don’t optimize for the technical integration first. Optimize for the seller incentive on the other side of the table. If the platform’s own sales team doesn’t have a reason to bring you into a deal, the listing doesn’t matter how good the API is.
The integrations that accelerate adoption are distribution plays, not engineering plays. Most teams get that backwards.
To keep the organization aligned before an AI launch, what single ritual or document do you use to get product, marketing, sales, and risk teams on the same page about what the system can and cannot do?
The document I use is what I call a capability boundary brief. One page. It answers three questions:
- What this system will do autonomously.
- What it will escalate to a human.
- What it will not touch under any circumstances.
The format matters less than the fact that product, legal, and sales have all signed off on the same answers before anyone demos the product externally or writes a press release.
The ritual I pair it with is a joint session where I put sales and risk in the same room at the same time—not separate reviews, not sequentially, together. Sales tends to overclaim capabilities in their enthusiasm. Risk tends to block things for hypothetical reasons. The conversation that happens when those two functions are in the same room is the most productive hour you’ll spend before a launch. Things that would have been arguments in separate emails get resolved in twenty minutes.
Most AI launch failures I’ve seen aren’t technical. They’re someone in the field saying the product does something it doesn’t do, because they sat through a great demo and filled in the gaps with optimism. The capability boundary brief is the antidote. You hand it to a new sales rep on day one and say: here’s what you can promise, and here’s exactly where you stop.
If the document can’t fit on one page, you don’t have alignment yet. You have a committee document that everyone will interpret differently.
One page. Three questions. Everyone signs it.
Thanks for sharing your knowledge and expertise. Is there anything else you'd like to add?
One thing I’d add: the conversations I find most useful right now aren’t about AI capabilities. They’re about what happens after deployment.
Most content covers how to evaluate AI, how to choose a vendor, and how to run a pilot. Very little covers what it takes to sustain an agentic system in production over time. Who owns it when the internal champion who drove the purchase gets promoted or leaves? What happens when a model update changes behavior in ways no one anticipated? How do you retrain the people whose jobs shifted, not just those whose jobs disappeared?
Those are the questions I spend most of my time on. They’re also the questions that determine whether an enterprise AI investment actually delivers or quietly gets sunset after eighteen months.
If any of this is useful context for your readers, I’m happy to go deeper on any of it. You can find more of my thinking at kubersharma.com.