Nobody Is Ripping Anything Out
The layer on top, and the work it creates
One of my partners sent me a reasonable request this week. We have a fundraising CRM, we are building a second one for portfolio companies, and we keep a third view for relationships. His note was the obvious one.
At some point I would like to just toggle between all three in one master CRM.
He said afterwards that he was speaking from a usability perspective, and he was right to. Three systems that hold overlapping people and do not talk to each other is a bad experience, and the instinct to consolidate is correct.
My answer was that we are going to solve it, and that building the master CRM he described is the one way I do not want to solve it.
That distinction turns out to be the whole of this piece, and it is the first genuinely good news in a series that has otherwise been a catalogue of things breaking.
The consolidation reflex is fifty years old
Every generation of enterprise software has hit this wall in the same place.
You buy the best tool for one job. Then the best tool for the next job. Five years later you have eleven systems, your people spend their days moving between interfaces, and the same customer exists in four of them with three different spellings.
At that point somebody proposes the suite. One vendor, one login, one data model. Rip out the point solutions and standardise.
I have implemented that project more times than I want to count. Portals, document management, CRM. It works, at enormous cost, and it works for about four years, at which point the suite is worse at each individual job than the specialist tools it replaced, and the cycle starts again.
The reason the cycle repeats is that consolidation has always meant consolidating the interface and the data in the same move. You could not have one without the other. If you wanted a single place to work, you had to move everything into a single system.
That constraint is gone, and almost nobody has updated their instincts for it.
What actually changes
Here is what I told my partner, and it is the thing I would put in front of any CIO with a stack problem.
An AI layer sits on top of all the applications, and you interact with the layer. The applications get relegated to an API.
That is the entire idea. You keep the specialist systems, because they are genuinely better at their jobs than any suite will be. You stop making people navigate them. The layer holds the context, knows which system owns which record, and does the work across all of them in response to a plain sentence.
I described a small version of this in part two. During an investment committee meeting I typed a sentence telling our platform to update a deal to reflect a meeting that weekend. It worked out what needed to change, which screen that lived on, made the change and asked me to confirm. My friend’s reaction was that I had just replaced six applications.
I had not replaced any of them. Every one of those systems was still there, still doing its job. What I had replaced was six user interfaces I used to have to learn and move between.
Consolidate the interface. Do not consolidate the systems. Those used to be the same project and they are not any more.
So my partner gets what he asked for, a single place to work across fundraising, portfolio and relationships. He does not get it by our building a master CRM. He gets it because the layer above all three stops caring which one holds the record.
For now we keep the individual portals, because there is still work that is faster to do directly, and because you do not migrate people to a new way of working by removing the old one on a Tuesday. Over time the centre of gravity moves to the layer, and the portals become the place you go when something needs a human hand.
Why this is a services opportunity and not a software one
Now the part that matters for the argument this series has been making.
Five articles have been about what AI takes away from the businesses that sell expertise. The invoice stops working. The pyramid loses its base. The catalogue gets absorbed by the platform vendor’s roadmap. If you sell professional services and you read all of that in one sitting, it reads like a wake.
This is the other side of the ledger, and I think it is larger than people expect.
Somebody has to do this work. An AI layer over a stack of point solutions is not a product you install. Every one of these is a different set of systems, a different data model, different permissions, different definitions of the same word, and a different set of decisions about what the machine is allowed to do without a person confirming it.
That is integration work. It is discovery, architecture, data mapping, orchestration and change management, which is to say it is exactly what systems integrators and consultancies already know how to do, pointed at a problem that did not exist three years ago.
And it is not the rip-and-replace project the industry has spent decades selling and clients have spent decades resenting. Nothing gets torn out. The point solutions stay, keep their licences, keep their contracts, and get demoted to an API with something intelligent on top.
That is a far easier conversation to have with a board than “we are replacing your stack.” The risk is lower, the disruption is lower, and the existing vendor relationships survive.
The work is not ripping anything out. It is relegating what you have and building the layer above it.
The prediction, and I will own it
Here is how I think this plays out, because it has played out this way every time.
It gets custom built first. Firms with the capability and the appetite build their own layer over their own stack, the way we are doing. Those builds will be expensive, uneven, and the good ones will produce a genuine advantage for a few years.
Then somebody builds the platform. A product that sits on top of the point solutions, that integrators can install and configure rather than construct from nothing. When that product arrives it will compress the custom market underneath it and create a much larger implementation market on top of it, which is precisely what happened with enterprise software and then again with SaaS.
The window between those two moments is where the money is, and it is open now.
If you are a services firm reading part two and wondering what you sell once you cannot sell hours, this is one honest answer. You sell the migration from a stack people navigate to a stack people talk to. It is a multi-year programme, it is genuinely hard, and the client cannot do it alone.
Why I think this is a waypoint and not a destination
Everything above is what I believe for now, and I want to be explicit that the phrase is doing real work. I do not think the layer is the end state. I think it is the step that makes the end state possible.
Follow the logic of relegation one move further and something uncomfortable appears. A system nobody opens any more is not really a product. It is a data model and a set of endpoints. You are paying enterprise software prices for an interface your people have stopped using and a feature set they were never going to touch.
At the same time, the cost of building software has collapsed. Not to zero, but far enough that the build-versus-buy calculation that has held since the 1990s stops returning the same answer.
So the question after this one is not which layer you buy. It is why you are renting eleven systems whose interfaces you no longer use.
The next target is not the interface. It is the point solutions themselves, replaced by custom, fully integrated software built by AI.
I am not claiming that happens next year, and there is a serious argument against it that I made myself in part four. The moment you build the software, you own it forever. You have taken on a maintenance obligation, a security surface and an integration burden that a vendor was previously carrying for a subscription fee. Software becoming cheap to write does not make it cheap to keep.
That tension is the next piece, and I do not yet know where I land on it.
The catch, which is the same catch as always
I want to be careful not to oversell this, because there is a version of it that fails.
A layer over a stack is only as good as the discipline underneath it. If the systems disagree about who the customer is, the layer will answer confidently and be wrong, which is worse than a person clicking through four screens and noticing the discrepancy.
And the governance question from part one becomes sharper rather than softer. When a person works in a CRM they can see what they are changing. When they type a sentence and a system decides which of eleven applications to write to, the confirmation step is no longer a nicety. In our own platform no AI writes a change to a record without a person confirming it, and that is not decoration. It is the thing that makes the layer safe to trust with more.
The firms that will win this work are the ones that treat it as an architecture and governance problem that happens to involve AI, rather than an AI problem that happens to involve architecture. Those are different projects and clients can tell the difference within about a month.
Where this leaves the series
Six articles in, the shape is clearer than when I started.
The economics of selling expertise by the hour are breaking. The labour pyramid that funded the industry’s growth is losing its base. The catalogue is exposed to platform roadmaps in a way almost nobody prices. Underneath all of it sits a question about what a firm keeps for human judgment.
And running alongside all of that destruction is a construction project of real size, which is moving every enterprise from a stack of applications people navigate to a layer people talk to.
That is what I believe for now. The step after it, replacing the relegated systems outright with software built for the purpose, is larger and less settled, and it is where this series goes next.
The firms that only read the first half of that sentence will spend the next three years defending a business model. The ones that read the second half will spend it building the thing that replaces it. At Bambu Capital that second group is what Services-Tech means, and it is what we are underwriting.
Nobody is ripping anything out.
They are relegating it, and building on top.
Somebody gets paid to do that.
For now.
A six part series on what AI is doing to professional services.
One. AI Decides How. Humans Decide What, Why, and If. Governing AI inside a single firm, and where the human veto has to sit.
Two. Nobody Is Buying Hours Anymore. What happens to an industry that priced and sold the one thing the machine turned out to be best at.
Three. Guess Who Becomes AI Native. Whether this wipes out a generation of young professionals, and what becomes of the pyramid.
Four. The Offerings Transform Too. The second wave, where the platform makes your catalogue unnecessary, and why productising is the wrong answer.
Five. Services-Tech. Naming the category, and the line between the services businesses that get rebuilt and the ones that do not.
Six. Nobody Is Ripping Anything Out. You are here. The AI layer over the stack, and why the migration to it is a very large services opportunity.