AI is forcing a fourth wave before the third finished.
June 8, 2026
In January 2023, I wrote a piece on what modern CFOs need to know about consolidation 3.0. The argument was straightforward: the legacy, on-premise, IT-owned platforms of the 1990s and 2000s had run their course. The third wave — cloud-first, finance-owned, no-code, faster to implement — was arriving, and vendors who didn’t adapt would be left behind.
Three years later, I stand by every word of it. The direction was right. And if I'm honest, the mid-wave reality of 2026 doesn't surprise me: consolidation vendors have rested on their laurels for too long, comfortable in the knowledge that consolidation is sticky and customers don't replace quickly. What I didn't anticipate was that AI would arrive and start the fourth wave before the third had barely begun. The promise of 3.0 — truly out-of-the-box, live in weeks, finance-owned from day one — has been partially delivered by some and barely started by others. The marketing caught up faster than the product did. And now the clock has reset, at a speed the market has never seen before.
I’ve lived through all four generations of this market. Consolidation 1.0 which was on-premise, IT-owned, and programming-heavy which gave way to Consolidation 2.0 roughly a decade later. It was another decade before 3.0 arrived. Each generational shift took ten to fifteen years to fully play out. Vendors had time to see it coming, debate it, and gradually reposition. The pace was evolutionary.
Consolidation 4.0 is not following that timeline. We are barely three years into 3.0 — and many vendors have yet to deliver on its promises. AI doesn’t move at the pace of previous technology waves. It moves at the speed of thought. The vendors who assumed they had the usual decade to adapt are going to find themselves facing two unsolved problems at once, on a timeline that gives them room to solve neither properly.
That’s the real disruption. Not just that AI is coming. But that it arrived while most of the market was still mid-journey on the previous wave.
For as long as I’ve been in this industry, there has been a comfortable assumption that mid-market consolidation is simpler than enterprise: Fewer entities, fewer users, and simpler requirements. It sounds reasonable, but it’s wrong — and it’s been wrong for years.
Mid-market finance organizations are carrying enterprise-level structural complexity on mid-market resources.
Think about what that actually means in practice. A company completes an acquisition mid-year. The platform handles the routine close. But the purchase price allocation, the goodwill calculation, the stepped acquisition adjustments — these almost universally end up in Excel, because they aren’t pre-configured in the application. The finance team posts a manual journal entry and moves on. The same is true for disposals, re-organizations, discontinued operations. Adding a new reporting currency or a new sub-consolidation level, something that should be routine, frequently requires configuration work that looks a lot like the coding that 3.0 was supposed to eliminate.
Mid-market companies have had no choice but to rely on these workarounds — not because their teams aren’t capable, but because no vendor gave them what they actually needed. Either the solutions that could handle the complexity were too expensive, took too long to implement, or were too dependent on IT and outside expertise to own and maintain. So, finance teams adapted. They built spreadsheet bridges, posted manual journal entries, and created parallel processes outside the system. And they called it normal.
The key difference between mid-market and enterprise is not complexity; its resources. A large enterprise has specialists—teams of people who manage the edge cases, run the parallel workstreams, and absorb the manual effort. A mid-market finance team often has a controller who is a one-person band, responsible for everything, with no one to hand the hard cases to. That controller doesn’t need a simpler product. They need a more complete one — a platform that handles what enterprise customers pay consultants to handle, without requiring a consultant to set it up.
That gap, between what 3.0 promised and what most vendors have actually delivered for this customer, is the unfinished business the fourth wave has to resolve.
Speed without confidence is just compressed risk. I’ve seen plenty of organizations close in five days and still not fully stand behind their numbers by the time they reach the board.
Every vendor in this market is talking about AI. Most of them are doing one of three things:
These are not without value. But none of them are what consolidation actually needs — and the gap between the AI being marketed, and the AI being built reveals how few vendors truly understand their customers.
Extended close cycles and a lack of confidence in the numbers share a common cause: errors that surface when the consolidation is run. But that’s not where they originated. Too many errors are still falling through the cracks upstream — in the trial balance import, the manual continuity schedule, the intercompany submission, the journal entry posting — silently accumulating until the consolidation engine exposes them. By then, the corporate controller has to trace them back to the entity controller, who investigates, corrects, sometimes goes all the way back to the ERP. The close cycle stretches. It’s the financial equivalent of chutes (snakes) and ladders — every data error is a chute that sends you back to the start.
AI strategically embedded upstream — where the data originates, before errors have a chance to propagate — is the ladder. Catch the problem at the source, and everything downstream improves automatically.
What consolidation needs is a data guardian: AI embedded throughout the process from the very first design decision, active at every data entry point, present in every entity submission workflow, monitoring from the moment a period opens. Not reacting to errors. Preventing them.
But there is a design principle that separates trustworthy AI from dangerous AI in this context — and it is one that most vendors promoting full automation have gotten wrong. Accountants are cautious, deliberate, and conservative. Not because they are resistant to technology — most finance professionals embrace AI enthusiastically. But because when it comes to the data itself, they are right to want control. These are the numbers a CFO signs off on. The numbers that go to the board, to auditors, to regulators. In consolidation, the numbers need to be right — submitting inaccurate numbers can be catastrophic, both professionally and legally. The instinct to retain control over them is not a limitation to be designed around. It is a professional and ethical responsibility.
The vendors pitching full automation — AI that acts on the data, posts the entries, accepts the eliminations without human review — have misread what consolidation accountants need from AI...
The right design is an agent that is autonomous on process tasks — sending reminders, enforcing submission gates, routing approvals, escalating overdue items — and suggestive on data decisions – diagnoses, prepares, and proposes. The controller reviews and decides. AI does the heavy lifting; the human is always in control. That boundary is not a limitation of the vision. It is the vision. And it is what makes AI in consolidation trustworthy rather than merely impressive in a demo.
But here is the harder truth: AI, even when designed correctly, is only as good as the data it operates on. And that points to a problem that none of the first three waves ever seriously addressed.
Waves 1.0 through 3.0 all optimized the same thing: the process layer. How fast you could close, how much IT involvement you needed, and how much coding was required. Each generation made the mechanics better. Not one of them fundamentally asked whether the data flowing through those mechanics was actually trustworthy.
Every generation inherited the same original sin: the consolidation engine was optimized; the data feeding it was not. The technology to enforce a genuine data trust framework across a multi-entity, multi-system finance organization simply didn’t exist at scale. So, the industry accepted an implicit assumption that errors would be caught downstream, manually, by controllers who had learned through hard experience where the landmines were buried.
That assumption is no longer acceptable. And with the right architecture, it is no longer necessary.
A data trust layer establishes a single certified data foundation — one integration per source system, one canonical definition for every financial metric, one validated version of the truth. Trial balance figures. FX rates. Journal entries. Intercompany balances. Ingested once, validated once, certified once. Not duplicated across applications with slightly different logic in each. Not reconciled manually every time the same number appears in a different report. Certified once. Trusted everywhere.
A genuine trust layer transforms the entire close and consolidation process. Data quality gates are enforced at every submission point, so that by the time the consolidation engine runs, the data has been certified at the source. Close and consolidation stop being two processes connected by workflow and become one integrated discipline unified by a certified foundation — one where the numbers are trusted before the consolidation runs, not reconciled after it finishes.
And for vendors with a suite of products — FP&A, Close, Consolidation, Account Reconciliation — the compounding effect is profound. The FX rate in the consolidation is the same certified rate in the FP&A model. The journal entry approved in the close is the same entry in the audit workpaper. One version of the truth, flowing everywhere. The CFO stops asking “why is this number different here than there?” — because the architecture makes it impossible for the number to differ. AI operating on siloed, inconsistently defined data is a liability dressed up in a convincing demo. AI operating across a unified, certified data foundation is a different proposition entirely. The combination of an AI data guardian and a trust layer is not an incremental improvement — it is the architecture that defines Consolidation 4.0.
I’ve spent thirty years at the center of this market. I know most of the mid-market players not as an analyst who evaluates them from the outside, but as someone who has worked at them, alongside them, and competed against them. When I started thinking seriously about what I wanted to do next, I evaluated the field carefully — including vendors I’d worked with closely, conversations with implementation partners, and a rigorous look at who was genuinely building something new versus repackaging what already existed under an AI narrative.
What I found, mostly, was the same pattern: 3.0 capabilities still being completed, AI planned as a future addition, and the data trust problem left entirely for another roadmap cycle.
What I found at Prophix was different — and it wasn’t a plan. It was already in motion.
Prophix has been a pioneer in applying AI to financial performance management long before AI became the industry’s favorite talking point. This isn’t a reaction to market pressure or a pivot to capture a trend. It is a continuation of a direction that was set years ago and is now accelerating — because the technology has finally caught up to the vision.
At the core is a consolidation engine that Prophix has been building for twenty-five years — and that depth matters more than it might appear. Consolidation is deterministic: there are right answers, and getting them requires an accumulated body of domain logic, built edge case by edge case, customer by customer, over decades. AI is probabilistic — extraordinarily powerful at detecting patterns, surfacing anomalies, and accelerating workflows, but reasoning in likelihoods rather than accounting certainties. AI alone cannot build in months what Prophix has built in twenty-five years. What it can do is make that foundation dramatically smarter. That is the combination: a robust battle-tested deterministic engine that guarantees correctness, with AI woven throughout to catch what humans miss, prevent what used to be caught too late, and reduce the manual effort that still consumes too much of every close cycle. A trust layer being built as the architectural foundation across the entire suite. AI embedded as a design principle from the first decision, not added when the build is finished.
That combination — a certified data foundation, AI throughout, across a suite that has been building toward this moment for years — is what Consolidation 4.0 actually looks like. Not a consolidation platform with AI features added on top. Not a trust layer bolted onto an existing architecture. A product designed from the inside out for the moment we are now in, by a company that saw it coming and has been building toward it ever since.
I also came here because of the leadership of the executive team. Alok Ajmera, CEO of Prophix, has a clear perspective: the vendors on the wrong side of the AI fence are going to lose, and this is the window to be on the right side of it. That’s not bravado — it’s an accurate read of where the market is heading, and it’s backed by the product decisions that have already been made. A company that has been defining AI in this space rather than reacting to it, and that is now doubling down at exactly the right moment, is exactly where I want to be.
