The question underneath the AI question
Finance organizations are adopting AI quickly. Far fewer report getting the value they expected from it. We have written about that gap before, in Finance Has Adopted AI. So Where Is the Value?
The usual explanations are about the AI itself. Wrong tool, wrong use case, needs better prompting, needs more time.
There is a less discussed possibility that I think explains more of it. Some AI initiatives underperform because of the financial foundation beneath the AI, not because the AI is incapable. The model is doing what it was asked to do. It was just asked to do it on top of information that was never actually aligned.
Which leads to a question most companies have not asked precisely enough.
Data flows everywhere. So why is Finance still reconciling?
Look at a typical growing SaaS company.
The ERP connects to billing. Billing connects to the CRM. The HRIS feeds the planning tool. The warehouse pulls from all of them. There are connectors, syncs, scheduled jobs, and a diagram somewhere showing arrows between every box.
By any technical standard, the systems are connected.
And Finance still spends the first stretch of every reporting cycle figuring out which number is right.
That is not a contradiction. It is the predictable result of confusing two different things.
Technical connectivity is not financial connectivity
Integrations move records between systems. That is what they are built to do, and modern ones do it well.
What an integration does not do is decide what those records mean.
It does not reconcile conflicting versions of the same fact. It does not standardize a definition across two systems that were designed with different assumptions. It does not resolve timing differences. It does not establish who owns a number when sources disagree. And it does not determine which version Finance should stand behind when the board asks.
Here is the principle I would put on the wall:
Connected systems move data. Governed financial infrastructure establishes what that data means. AI needs the second one.
Most companies have invested heavily in the first and assumed it produced the second. It does not, and the gap between them is where the reporting cycle goes.
What this looks like in practice
These are the disagreements that show up in real finance organizations, none of which indicate anything is broken.
Salesforce ARR does not match billing ARR. The CRM records what was sold, at the amount on the closed opportunity. Billing records what is actually being invoiced, after amendments, proration, and terms that changed during implementation. Both are correct. They are measuring different moments.
Billing activity does not cleanly reconcile to revenue or the GL. Invoicing and recognition follow different rules. Deferred revenue, ASC 606 treatment, credit memos, and annual billing on monthly service all create legitimate differences that someone has to explain rather than resolve.
Forecast definitions differ from reported actuals. The forecast was built on bookings because that is what the sales motion produces. The actuals are reported on recognized revenue because that is what accounting produces. Comparing them requires a bridge that usually lives in one analyst's workbook.
Sales and Finance define churn differently. Sales counts logos at renewal date. Finance counts dollars at period end. Both numbers get presented, sometimes in the same meeting, and the discrepancy gets explained verbally every time.
Headcount classifications do not match. The HRIS counts employees by employment status. The planning model counts budgeted requisitions including open roles and contractors. Neither is wrong for its purpose.
Reports use different cut-offs. One pulled Tuesday, one pulled Friday, one after late invoices landed. Same metric, three values, no error anywhere.
The failure is not the software
I want to be clear about this, because the instinct is to blame a system.
Each of those systems is doing exactly what it was designed to do. The CRM is supposed to track the sales process. The billing platform is supposed to invoice correctly. The HRIS is supposed to manage employment records. None of them was built to produce a company's financial truth, and it is unreasonable to expect that they would.
The gap is not inside any system. It is in the space between them, and nobody owns that space by default.
So Finance fills it. Every period, someone pulls the data, lines the systems up, applies the company's definitions, resolves the timing, decides which version wins, and produces one coherent picture. The integration layer in most growing companies is a person with a spreadsheet and a deadline.
What Finance actually needs beyond synced records
Synced records are the starting point, not the destination. The financial foundation requires several things that no integration provides on its own.
Common definitions, so expansion means one thing in the board deck and the pipeline review alike.
Reconciliation, so differences between sources are identified and explained rather than quietly averaged.
Validation, so incomplete or anomalous data surfaces loudly instead of flowing through into a report.
Timing consistency, so figures compared against each other were pulled on the same basis as of the same moment.
Ownership, so it is clear which source is authoritative for which fact.
Traceability, so any number can be followed back to the calculation and the source record behind it.
Controlled change, so when a definition is updated, it is updated deliberately and everyone downstream knows.
Approval of final outputs, so a person signs off before anything is published.
That is what I mean by governed financial context. It is not a database project. It is the set of agreements and controls that turn moving data into a number someone can defend.
Five tests of whether Finance is actually AI-ready
If you want a practical read on where your organization stands, these five questions get you most of the way there. They are diagnostic rather than rhetorical, and the honest answers tend to be uncomfortable.
Test 1: Do your systems agree?
Can Finance explain the differences between ERP, CRM, billing, HRIS, and planning without rebuilding the reconciliation manually every period?
Explaining the difference is fine and often correct. Rebuilding the explanation from scratch each month is the signal.
Test 2: Do your definitions agree?
Do ARR, MRR, churn, expansion, customer, bookings, revenue, and headcount mean the same thing across systems and reports?
A useful version of this test: ask three people in three functions to define net revenue retention, and see whether you get one answer.
Test 3: Can important numbers be traced?
Take a figure from the last board deck. Can you get from that number back to the calculation that produced it, and from there to the source records underneath it, without a conversation?
If the answer depends on who is available, the number is available but not defensible.
Test 4: Can the process run without one person holding it together?
Look for spreadsheet dependencies, manual exports, undocumented overrides, and tribal knowledge. Then ask what happens during close week if that person is unreachable.
This one is rarely a technology problem. It is usually an unfunded continuity problem.
Test 5: Does AI operate after financial truth is established?
Is AI analyzing numbers that deterministic rules and validation already produced? Or is it being asked to work out what the numbers are?
The first is a capacity multiplier. The second is a risk.
Why AI raises the stakes on all of this
Every one of these problems existed before AI. Finance organizations have absorbed them for years, largely through effort.
AI changes the consequences in a specific way. It can synthesize inconsistent information convincingly.
A spreadsheet with a broken link often looks broken. An AI-generated explanation of misaligned data looks like a well-written analysis. It reads confidently, uses the right vocabulary, and gives no visual indication that the ARR figure underneath it came from a source that disagrees with billing by four percent.
Hidden data problems used to surface as friction. Now they can surface as fluent, plausible, wrong narrative delivered to an executive audience. That is a meaningfully worse failure mode, and it is why foundation work that was merely valuable before has become a prerequisite.
This is not an argument for one giant database
To be clear about what I am not saying.
Finance does not need every operational system consolidated into a single database, and it does not need all systems to hold identical data. That project is expensive, slow, and usually fails, and it fights the reason those systems exist. The CRM should keep tracking the sales process the way sales needs it tracked.
What Finance needs is a governed financial context that sits across those systems. One place where the data is reconciled, the definitions are enforced, the timing is consistent, and the numbers are produced the same way every period. The source systems stay exactly as they are.
That layer is what a Finance operating system is meant to provide, and it is the difference between systems that are connected and financial information that is aligned.
Reframing AI readiness
Most AI readiness conversations in Finance are really procurement conversations. Which tools, how many seats, which use cases first.
I would propose a different measure.
AI readiness is not how many AI tools Finance has deployed. It is whether those tools can operate on financial information the organization actually trusts.
By that standard, a company with one AI capability sitting on a governed financial foundation is further along than a company with six sitting on integrations alone. The first will produce answers people can act on. The second will produce more answers, faster, with no way to tell which ones are right.
Connectivity was the last decade's project. Financial alignment is this one's.
Related reading
Earlier in this thread: Finance has adopted AI — so where is the value?. Next: financial data validation. Cornerstone: AI operating system for SaaS finance.
SMPL.ai is an AI-powered financial intelligence platform for growing SaaS companies. It brings financial and operational data together across systems so Finance can establish trusted financial context before using AI for analysis. Learn more at www.smpl-ai.com.