Don’t Bet Your AI Strategy on Today’s Winning Model

A year ago, if you asked which company was winning the AI race, you probably had an answer.

Ask the same question today and you might give a different one.

Ask it a year from now, and I suspect the answer will change again.

That’s why an article from CreativeCo Capital caught my attention this weekend. The article explores what it means for software to be truly “AI-native,” but one point in particular stood out to me: AI architecture should be model agnostic.

Different models are already better suited to different tasks. Their capabilities are changing rapidly. So are their economics. CreativeCo argues that the enduring value should therefore live in the proprietary data, tools and context surrounding the model, not in a long-term bet on any single model provider.

That raises a question I think more businesses should be asking:

Are you building an AI strategy or accidentally making a long-term bet on today’s AI leader?

The horse race is just getting started

I sometimes describe the current AI market as a horse race.

The lead keeps changing.

New models arrive. Existing models improve. Prices change. One model may be better at one task while another excels somewhere else.

That competition is great for businesses. But it also means your AI strategy shouldn’t depend on correctly predicting the winner.

Your data should be able to move with the frontier.

That was an important part of the presentation we shared at PDI Connections last week.

Over the last several years, our work with retailers has expanded from connecting transactional data to adding fuel, labor, loyalty, inventory, waste, site information and, most recently, General Ledger data.

The questions have evolved along with it - from better reporting, to understanding what happened around the transaction, to connecting operational activity with business outcomes.

But this year we demonstrated something new:

What happens when you make that connected data foundation available to the AI platform the retailer chooses?

Own the foundation. Let the models compete.

At Connections, we debuted Taiga’s new MCP capability.

MCP - Model Context Protocol - allows us to make the connected, cleaned and contextualized data in Taiga securely available to AI tools.

Instead of beginning with a predefined report or dashboard, a user can begin with a question:

• Which stores need my attention today?
• Why is this store underperforming?
• Sales are up, but are we actually making more money?

The AI can work across the business context available through Taiga to help answer those questions.

But to me, one of the most important things about this architecture is optionality.

Today, you might want to use ChatGPT. Another company may prefer Claude or Gemini. Tomorrow, the best choice may be something entirely different.

We don’t think retailers should have to rebuild their data foundation every time the AI leader changes.

Own the data foundation. Preserve the business context. Let the models compete for the opportunity to use it.

Look beyond the chatbot

Over the next year, nearly every enterprise technology provider will have an AI story.

Many of the demonstrations will be impressive.

But I think businesses should look beyond the chat window and ask a few questions:

• What data can the AI actually access?
• Does that extend beyond one vendor’s application?
• Does it understand the relationships across my business?
• Can I use that same data foundation with another AI platform?
• If the AI landscape looks completely different two years from now, how much do I have to rebuild?

Those questions may ultimately matter more than which model produces the most impressive demonstration today.

Don’t make your data dependent on your AI. Make your AI dependent on your data.

At Connections, we ended our presentation with a simple thought:

Your questions will change.
Your technology will change.
AI will change.
Your data foundation needs to be ready for all of it.

Nobody knows which horse is going to win the AI race.

The good news is that businesses don’t need to know.

Build a data foundation that understands your business. Keep control of it. Then let the AI models keep racing.