Predictions for the closed-loop AI market.

Large scale generative AI has dominated the headlines for what seems an eternity already. But developments in closed-loop AI - business applications restricted to using the data within an organisation - is where I think the interesting stuff is happening.

Here’s a ‘from the trenches’ view of the current state of play in this market from friends, colleagues and clients in the business.

Overnight, every CEO felt the pressure to prove they have an AI strategy in place. Of course, all their suppliers pivoted to say yes, they could help.

Usually, the first contenders were the consultants. They’d suggest the business processes ripe for streamlining and recommend some AI tools that could do the job.

Then the baton was passed to the IT developers and integrators. They’d configure a system with various business processes and plug in all the data sources.

The holes in data quality quickly became apparent when the system was tested. Without the right data, the system was little more than an idiot savant. So, the baton was finally passed to the data consultants to clean things up.

At some point, (in many cases too late), the board started to think about AI governance. If something goes spectacularly wrong with the system, who’s culpable? It’s clear that the system needed management tools to provide oversight, audits and governance mechanisms.

The push then started to programme the system to reveal the insides of it’s black-box. The code needed to detail where the data came from, and the logic it applied for each process and decision.

Apparently the first part, is fairly easy, the second is trickier. The solution might be to install a separate AI to query the main AI to get the answers!

Then, once you have all the relevant data, what are the business protocols for management oversight, decision making, issue escalation, audits and compliance reporting?

And these issues are only set to become more complex. At the moment, these systems are mainly prototypes developed in silos for individual applications. But what happens when you want them to talk to each other? What does that architecture look like?

We’re only just beginning to tackle these challenges.

So, broadly speaking, that seems to be the current state of play. A patchwork quilt of suppliers, pivoting what they’ve previously done and applying it to the new frontier. All of them learning on the job.

Inevitably things will start to consolidate. I suspect we’ll see a slew of IT companies and system integrators scaling up specialist AI teams by acquiring or building out their data capabilities. Conversely, data consultants will scale up or acquire their dev teams.

This will certainly make the development process more efficient, and the thinking more connected. It will quickly become table-stakes for any development partner in the market.

However, my prediction will be those who offer a complete top-to-bottom solution will ultimately win the game. Teams comprising of business analysts, management consultants, system developers and integrators, data and governance specialists all working together with a 360 degree, connected mindset from the start.

Why?

Because plugging in Artifical Intelligence without the wisdom of how it impacts the business, management processes and governance is a dangerous thing.

And with the fractured, patchwork quilt of players on the production line at the moment, the danger is things will fall between the cracks.