We are an independent enterprise research body. Adoption is the whole subject.
Global AI Forum exists to answer one question properly: what is actually working inside enterprises that have deployed AI, and how would anyone know. We work with leadership teams in the room, and we keep a register of verified deployments underneath that work so the advice rests on evidence rather than on anybody's marketing.
Paid by the enterprise, never by the outcome
We take no referral fees and no sponsorship from anyone selling AI.
San Francisco · Chennai
Why I started the Forum
Co-chair, AI Council, Southern India Chamber of Commerce and Industry
I have spent most of my working life on both sides of the enterprise technology table. I ran business for a large services firm, I held senior roles inside a bank, and I have spent the last few years helping technology companies win enterprise customers. I have sat in the room when the deal was signed and I have sat in the room eighteen months later when somebody quietly asked what we actually got.
Those two rooms almost never agree. The first one runs on a story. The second one runs on whatever is left of it. The gap between them is not dishonesty, and I want to be careful about that, because most of the people selling AI into enterprises believe what they are saying. The gap exists because nobody is checking, and in a market where nobody is checking, the most confident claim wins by default.
The enterprise AI market has no shortage of opinion. What it is missing is somebody willing to go and ask the customer whether the thing actually works.
So that is what we built. We take a vendor's published customer claims, we write to the enterprises named in them, and we ask simple objective questions. Do you work with this company. To what extent. Is it running, and how far has it gone. Then we publish what came back, including when the answer was no, including when nobody answered at all. The enterprise is never named. The record is.
We charge vendors for that check, and I want to say that plainly rather than hide behind the word independent. A vendor pays us to examine its claims. It cannot pay us for the result, it cannot see the determination before it publishes, and who commissioned the work is printed on the record itself. If we ever find ourselves softening a finding because of who paid for it, the whole thing is worthless and we should stop.
The work I enjoy most is still the work in the room. Taking a leadership team through where they actually are, not where the deck says they are, and leaving with three decisions and the names of the people who own them. The register exists so that when I am in that room, I am bringing evidence rather than opinion. That is the entire design.
If you are an enterprise trying to make a decision you cannot unmake, or a company that believes its claims will survive being checked, I would like to hear from you.
What we are for, and where it ends
Two sentences we are willing to be held to. Everything else on this site should be traceable back to one of them.
Make enterprise AI adoption measurable, and make the measurement public.
Not forecast it, not rank it, not predict it. Measure what has actually been deployed, verify it with the organization that deployed it, and publish the result where a buyer can reach it. Everything we sell, from a workshop to a category review, runs on that same evidence.
A market where no enterprise AI decision rests on a claim nobody checked.
Where a buyer can see what is real before they commit a budget and a career to it. Where a company with genuine deployments has a way to prove it that does not depend on being the loudest. We will know we got there when being checked is normal rather than remarkable.
Five rules, and what each one costs us
Values are only worth printing if they are expensive. Each of these has turned down revenue, and we would rather say which.
Evidence over narrative
If we cannot show where a number came from, we do not print it. No estimates dressed as findings, no survey responses presented as deployments.
Say the number, including zero
Where a category has no confirmed deployment, we print a zero rather than a hedge. The empty rows in our tables are the ones doing the most work.
Independence you can audit
Who commissioned a piece of work is printed on it. Determinations publish on our cycle, not the vendor's, so no result can be timed to a funding round.
Refuse the easy money
No referral fees, no introductions sold, no sponsorship from anyone selling AI into the enterprises we advise. Several of those were offered. All of them were declined.
Publish what we could not answer
Unanswered enquiries and lapsed determinations stay visible. A method that only shows its successes is a marketing instrument, not a method.
The next three years, stated so you can hold us to it
We would rather publish a plan and miss part of it than describe a direction vaguely enough to always be right.
Depth before breadth
Nine industries and the first business functions carried to the same standard, with every claim traced to the organization named in it. The Enterprise AI 100 published annually. The method itself documented in public so it can be argued with rather than trusted.
The register becomes searchable by anyone
Every record reachable, every determination carrying its own date and its own commission line. Geographic coverage extended properly beyond North America and Western Europe, because the current skew is a limitation of ours and not a fact about the world.
An ombuds route with teeth
A published way for a vendor or an enterprise to contest a determination, heard by someone who did not make it, with the outcome printed whichever way it goes. A register that cannot be appealed is a register that will eventually be wrong in public.
Make ourselves ordinary
The goal is not to become the only body that checks. It is for checking to become a normal part of how enterprises buy AI, done by several credible institutions to a comparable standard. If that happens and we are one of four, we will have succeeded.
Why any of this matters
Enterprises are making decade-long decisions on evidence that would not survive a single phone call to the customer named in the slide.
The cost is not the license
A wrong AI decision rarely bankrupts anyone. It does something slower and worse. It consumes two years, a budget, the credibility of the person who championed it, and the organization's appetite to try again. The second attempt is always harder than the first.
The good companies are losing too
A market that cannot tell a real deployment from a confident one does not reward the companies with real deployments. It rewards the best storytellers. The firms with genuinely working products have the most to gain from being checked, and they are usually the first to say yes.
Somebody has to be paid by neither side
Analysts take vendor money. Consultancies sell the implementation. Integrators sell the license. Each is useful and each has a stake in the answer. We took the one position nobody wanted, which is to be paid for the check and never for the result.
Four ways in, and who each one is for
Three of these are for enterprises. One is for the companies selling to them, and it buys nothing but our attention for thirty minutes.
Expert guidance
For enterprises choosing between vendors or reading a category. A standing line to the people who maintain the register, bought as sessions rather than seats.
For enterprisesExecutive workshops
For a leadership team, a single function or the management layer. Your own AI maturity worked through against our frameworks and canvas, with decisions made in the room.
For enterprisesEvents
Convened rooms where enterprise leaders compare what is actually running, under rules that keep the selling out of it.
For enterprisesBriefing sessions
For vendors. Thirty minutes to brief us on what you do, free, on a monthly theme, with limited slots. It buys nothing except that we understand you properly.
For vendorsGet indexed
For companies selling AI into enterprises. We examine your published customer claims and publish what we find. You pay for the check. You never pay for the outcome.
Request a vendor
Anyone can ask us to open a record on a company that is not yet in the register. It is free, and it is usually complete inside ten working days.
Two places, on purpose
The questions an enterprise asks in San Francisco and the questions it asks in Chennai are not the same question, and a body that only hears one of them will get adoption wrong.
720 Market Street, San Francisco, CA 94102
+1 650 275 3007 · research@gaiforum.com
Sakthi Towers, Anna Salai, Thousand Lights, Chennai 600002
research@gaiforum.com
The register skews to North America and Western Europe, and we say so on every page that shows it. That is a limitation of our coverage rather than a finding about the world. Correcting it is the single largest piece of work in front of us, and it is why the second office exists.
Tell us what you are deciding
A shortlist, a category you are entering, a program that stopped moving, or a room you want us to run. We will tell you what the register can answer, what it cannot, and which of our services fits. If none of them does, we will say that.
What people ask about us
A vendor pays us to examine its claims, the same way a company pays an auditor. It does not pay for the outcome, it does not see the determination before it publishes, and who commissioned the work is printed on the record. Determinations publish on our cycle rather than theirs, so no finding can be timed to a funding round. If you think that arrangement is still compromised, the method is published and the ombuds route exists for exactly that argument.
A research body, and the distinction matters. We do not implement, we do not resell, and we do not take a position in anything we write about. We advise enterprises and we run rooms with their leadership teams, but the work is always drawn from the register rather than from a practice we are trying to staff.
Small, deliberately, and we would rather say so than imply otherwise. A small team that does the checking itself is more useful than a large one that subcontracts it. What matters for your decision is not our headcount, it is whether the determination in front of you names its source and carries its date. Both of those are on every record.
Yes. The offices are in San Francisco and Chennai, and the work runs across North America, Europe, the Gulf and Southeast Asia. The register's coverage is still weighted to North America and Western Europe, which we state wherever the numbers appear rather than letting a skew read as a finding.
Yes, on one condition: we keep editorial control and the result publishes whatever it says. A category review is commissioned work and it publishes like any other determination, with the commission line on it. What we will not do is write to a conclusion you have already chosen, however the brief is worded.
The Forum publishes under its own name and the editions carry a byline. The enquiries behind a determination are made by the people who maintain the register, and the organizations that answer do so under anonymity, which is why you will see a named vendor and an unnamed customer on every record. That asymmetry is deliberate and it is the only reason enterprises answer us at all.
Bring us the decision you cannot unmake
Check what the register holds on the vendors in front of you, commission a review, or put the decision in a room with your own leadership team.