We let AI write for us because it is inventive. We do not let it be responsible for anything.
That gap is not a technical problem waiting on a better model. It is the whole opportunity.
What they built
Veriad builds marketing agents for large enterprises.
The founders are Rohan Mahendraker, chief executive, a self-taught software engineer who has built and sold compliance software to Fortune 500 companies, and Anton Muratov, co-founder and chief commercial officer, who previously closed seven-figure deals with the largest bank in the United States and the largest fashion group in the world, advised Fortune 500s on regulatory policy, and before Veriad advised early-stage startups and nine-figure scale-ups on go-to-market. In a previous life he was a competitive long-distance runner.
Read those two together and the company is already visible. One has sold compliance software into the biggest companies in the world. The other has sold into banks and advised on what regulators will accept. Neither of them is a marketing person.
Who else is in this field
Marketing software is possibly the most crowded category in all of business. HubSpot, Salesforce Marketing Cloud and Adobe Experience Cloud own the enterprise stack. Marketo and Braze run the campaigns. Jasper and Copy.ai arrived early on AI-written copy, and Writer went specifically after large regulated companies. Typeface and Persado work on staying on brand. Contentsquare and Amplitude measure what happened afterwards.
Almost all of them compete on output: more content, better content, faster. That is the creative axis, and it is full.
The point: correctness is a responsibility, not a measurement
Here is the argument, and it is worth taking slowly.
People adopted AI for its inventiveness. It surprises us, it produces ten good options in a minute, and nobody minds that it occasionally writes something silly, because a silly first draft costs nothing.
Being right is a completely different matter, and the reason is not that machines are bad at it. It is that in professional life, being right has never been a property of the work. It is a position somebody occupies.
An audit is not true. It is signed. A legal opinion is not correct. It is given by somebody who can be sued for it. A drug claim is not accurate. It is approved, by a named person, against a written standard. In every one of these, the thing being bought is not the sentence. It is that a qualified human has put themselves behind the sentence.
A great many experts hold their jobs because they can carry responsibility, not because they know things nobody else knows. Correctness has always been an answer to the question of who is accountable when it turns out to be wrong.
Which is exactly why AI cannot simply walk into that role. An agent cannot be struck off, cannot be fined, cannot be called in front of a regulator, and cannot be sacked in a way that means anything. So creativity transferred to machines almost overnight, and responsibility has not moved an inch.
That leaves an enormous space. Everybody is competing on the creative side, which is crowded and getting cheaper every month. Almost nobody is competing on the other side, which is where the money and the fear both live.
What is actually being sold is defensibility
Follow the logic and the product stops being better writing.
What a bank needs is not prose that is more accurate in some abstract sense. It needs material it can defend: every claim traceable to a source, every disclosure present, a record of who approved what and when, and a process it can show a regulator afterwards. If all of that exists, the work is correct in the only sense the institution cares about.
So the thing worth building is not a more accurate writer. It is the apparatus of accountability around the writing, which is precisely what every law firm and audit house has always sold.
That is a much better business than better copy, for an obvious reason. Copy quality improves for everybody every time the model vendors ship, so nobody keeps an advantage. The ability to stand behind the output does not arrive in a model release. It is built, argued for with regulators, and proven over time.
The race is for the seat, and it is a short race
Now the part that makes this urgent rather than merely interesting.
Trust does not get re-examined. Once an institution decides a supplier is the one it can rely on for this kind of work, it stops asking. The decision is expensive to make and nearly free to repeat, so it gets repeated by default, often for a decade or more.
This is why the incumbents in every profession look unassailable and are mostly not better at the work. Nobody re-runs the comparison. Being reliable is not something a customer re-measures each quarter, and reputations are formed early and then simply persist.
So the prize here is not being the most accurate. It is being the first to be treated as the one you can rely on, because that position is granted once and rarely revisited.
Which makes this a land grab with a closing window. The companies competing today are competing on output. Whoever moves first to compete on accountability takes a seat that the others will find very difficult to argue anyone out of, long after their writing has caught up.
A caution on the same thought, since it cuts both ways. A position built on being trusted is lost catastrophically rather than gradually. One material failure in front of a regulator does more damage than years of good work repaired. Anyone taking this seat has to be genuinely able to hold it, because the thing that makes it valuable is the same thing that makes it fragile.
Why it is bigger than it sounds
The short description is marketing agents. The industry it points at is not a marketing industry.
- Correctness is a job, not a property. Experts are paid to carry responsibility as much as to know things, and an AI cannot be struck off, fined or held accountable. So creativity moved to machines and responsibility did not.
- What gets sold is defensibility. Not prose that is more accurate in the abstract, but material an institution can stand behind, which is what professional firms have always really sold.
- Trust is granted once. Nobody re-runs the comparison, so the race is to be first into the seat, not best in it. That window closes.
What to watch
The number that matters is the time from brief to published inside a regulated customer, because that is the step where the institution's own accountability process happens. Volume of content produced is a vanity figure in this market.
The second is whether any customer lets output reach a real audience substantially as generated. There is a large difference between a tool that produces a draft a person rewrites, and one the institution is willing to stand behind. Only the second is the business described here.
The third is the question to put to the founders, and the most revealing one. When the material is wrong, who is accountable? If the honest answer is still the customer's own compliance team, Veriad is a good tool inside somebody else's process. If Veriad is prepared to take a share of that responsibility, it has stopped being a marketing company and has started building the thing this file is actually about.