Snail-paced enterprise and Warp-speed AI

Perception of progress is progress, they say. I will twist it to say that ‘perception of progress’ leads to progress, in enterprises.

Enterprises are slow not for convoluted reasons. Friction and slowness ensure stability. They make sure that manageable risks don’t go out of hands. For modern AI companies encountering the enterprise for the first time, an enterprise’s ‘quarterly cycle’ of progress measurement might feel too slow. The enterprise inertia will stay and it is even important.

In AI, on the other hand, speed is the only moat. Anyone who can build software will build it. The moat comes from speed, data, and distribution. The more you work with an enterprise the more data you get and more distribution you are guaranteed for your products/services.

Most AI folks who work with an enterprise assume that signing the contract is the main battle. No. That’s just the right of passage. The real battle is getting everyone to make the decisions you want them to, even if they all agree on a common goal.

Whether you are a ‘founder mode’ executive in an enterprise or a founder selling to an enterprise, I have seen a few things work. You may just find them useful in your situation too:

1. Don’t ask for data

   Its a muscle-memory for enterprises to say ‘No’ to data access requests. The ‘No’ may come wrapped in security questionnaires, meetings over Microsoft Teams (which is a hostile tool that all enterprises put in front of you to check your commitment to serve them), and occassionally a steering committee meeting where you wear a sports coat to, and eat cold turkey sandwiches. My point is, asking for data is the best way to be denied data. Create synthetic data. Show how AI works. Negotiate for access. Don’t assume it will be available on day 1.

2. Make friends in the business team

   Your enterprise’s business users are your best friends. Its likely that they are bogged down by how slow decision-making is. Serve their needs. Build demos for them. Make them want it. They will eat cold turkey and speak for you, over countless Teams meetings.

3. Listen to the nay-sayers

   A nay-sayer chewing turkey sandwich between their ‘No’s’ is not your enemy. They are the manifestation of the enterprise machinery that runs at its pace. They represent the reasons and concerns that if you don’t address, you may not win. Business will work with you, share requirements, explain the opportunity, and even demo your work in every meeting but they can’t get your effort to production. Every production-izing conversation starts with a bunch of gasps, lumbar support creaks of chairs being stretched back by the CIO team that literally wants to pull themselves back from the hot mess you are creating. They know what other hot messes exist. They don’t have time. Find a friend. Find a translator. Find another CXO that can mediate. There is just no way you launch without checking the boxes on production readiness. Switch from demo mode to deliberate mode.

4. Don’t under-estimate value

   As a creator, you may live in the future. An LLM-based dashboard may be trivial to you but it would be of immense value to the business. It’s the tangible outcome that the executive board can feel. It might be a two-week endeavor but it is more important than the copilot that you are building. When the enterprise gets excited, embrace that path.

5. Be Resourceful

   Don’t wait for ETL pipeline work to begin. It never might. Find a contractor to introduce. Join your client in a trade show to show demos. Let the pull from the market create urgency. Look at the RFPs your client is participating in. See how your product can give them an edge. Put out a demo that can go with that RFP. Plant your MLOps engineer with the IT team for a couple of days of fact-finding. Be seen as enablers and not disruptors.

6. Live with the dichotomy

   You will be asked to talk about the ‘latest’ in AI and even demo it, while being asked to go slow. The enterprise wants to embrace the latest. It’s pains are real. But the org debt is real too and that means no new features can go live as fast as the demos do. Yet, you have to know the latest about models, text-to-speech, knowledge graphs while being comfortable with months to launch anything

We are yet not at the ‘No one gets fired by implementing IBM’ phase of enterprise AI deployment. But we surely are at orchestration of data, quick/low-hanging isolated wins like document extraction, chatbots, and LLM dashboards.

The smartest way to enable an enterprise on their AI roadmap is not to say it will take 18 months and a Phase ‘0’ of preparation, but by showing what’s possible, embracing the shortest routes to value, and showing the IT teams that you will work with their aspirations and anxieties.