Buying vs. Becoming AI

The part of my work as the rainmaker for Moative I love the most is talking to CXOs in industries that have been around for several centuries and are cautious and conservative about adopting tech. The part that I hate the most is the thing they pass off as Biscoff in airline snacks.

I must confess that tech has largely been useless in unlocking labor productivity for industries for whom tech is CNC machines and boilers. The “software tech” gravy train lead by SAP, Oracle, and the likes largely just meant more work, proof of work, and record-keeping for compliance. People worked for these software tools than the other way around. So I understand the trepidation towards software buying.

This tantalising promise trap

I am sure that you have used and been using ChatGPT. It mostly works. You see magic and wonder ‘What if, we can ask ChatGPT some questions about our parts catalog?’. And then you see it hallucinating. But you see another ‘blow your mind’ use case being discussed on LinkedIn. So you do what every executive with 25 years of experience does – go to a conference and get mobbed by AI vendors and thought leaders.

You come back invigorated and nervous.

It’s a trap.

The truth is AI models continue to hallucinate. Products built on top of AI models are still waiting for rigorous evaluation frameworks (in plain speak, tools that prevent their AI from calling itself ‘MechaHilter.’ Look up, if the term hasn’t come to your radar yet).

It is also true that SaaS vendors are dogmatic and they don’t understand data as much as they understand software technology. It’s the same as someone understanding milling but not metallurgy.

So to account for the lack of understanding of how good (or bad) your data house order is and what it takes to clean, glean, and run their AI on your data, they slap a set up fee. That fee is simply “cover our back side” price.

The dogma of SaaS pricing best practices calls for extractive ideas like recurring billing, staggered plans each locking you out of what’s needed and push you to the next plan. I will explain the trap now.

SaaS is not creating the value

The industry-vertical SaaS that pitches to you in your industry language – what are they pitching? Smart extraction of data from documents; reconciliation of records across systems; autonomous prescriptions of decisions to take; and in some cases, take that action.

It’s all impressive but none of these capabilities are vendor-locked. Extraction today costs 7c per 1000 pages. There – I spilt the economics for you. Reconciliation, prescriptions, and decisions are all AI models that the SaaS companies have stitched together.

The value is in cleaning your data, stitching the models to the workflows, and evaluating/setting guard-rails.

There is value but is it worth the recurring software bills? If AI models are getting cheaper and cheaper, is your vendor going to reduce the price next year?

Consultants are pickpockets

I know, I know. It’s a bit harsh. Accenture, Cap Gemini, BCG Consultants do not put a long drawn execution plan and milk you for 7 figures. They totally don’t mind if implementing AI agents means less work for their consultants.

My point is you need AI-first consultants that don’t bill based on hours or at least, have a very different pace to execution, thanks to the AI models they use for software engineering and their pragmatic approach to stitching models using tools like Make, Gumloop, and n8n.

If your consultant does not do that, read the sub-heading again.

Why build AI when you can become the Industry AI

Honestly, the real value in the AI game is your data and your org/industry context. The tools, the models, the consultants all depend on that. So why trade that value for nothing?

As far as I know, Moative is the only company that is neither a product company nor a consulting company. This is a ‘once in a century’ opportunity for you – the distributor, the paper mill, the pipe maker – to co-build an industry-first AI, grounded on your industry context.

If AI is good, and your context makes it powerful, should you not build it yourself and get the upside by making it a product for others to use? Should you not become the “Industry AI” company?

What comes in the way? I raked my brain about it. The only thing that comes in the way is dogmatic SaaS vendors and conservative consulting firms that won’t trade their ‘today’s profit’ for ‘tomorrow’s industry leadership.’

We will.

I know, this sounds like a marketing flyer slipped under your door by the paper boy. But think about it. Why should a paper mill share 50 years of their data to get no IP for that AI and instead pay 6-figures to someone who stitched someone else’s AI. Is integration so valuable that you leave IP value at the table? Yes, you own your data as per the contract but someone else is building a dominant AI platform without any understanding of the industry, until you help them.

Previously software engineering was the value that justified their dominance. Now, that value has moved to general purpose models. Coding has become a commodity. The only thing that matters is your context. Contracts keep your data safe but the industry context (workflows, decisions, exceptions) are all not protected.

Would it not be easier for you to co-opt with an AI company to launch an industry-focused AI venture? You have the distribution already. You have the know-how. Now, you can show the way.

Enough caffeine for Sunday?