The Finance Team’s New Superpower Isn’t a Dashboard — It’s a Conversation about Hypotheses
by Don Dittmar
I’ve spent the better part of two decades building products that sit between finance teams and their data. And for most of that time, the fundamental constraint has been the same: you can only find what you know to look for.
That’s changing. Fast.
If you’re in finance operations and you haven’t encountered the Model Context Protocol (MCP) yet, you will. And when you do, it’s going to reshape how you think about what your role actually is.
The Old World: Structured Queries for Structured Questions
Here’s how anomaly detection has worked in AP and finance ops for essentially forever: someone builds a rule such as:.
- Flag any invoice over $10,000.
- Alert me if a vendor submits twice in 30 days.
- Show me variance against budget greater than 15%.
These rules work. They catch the things you already suspected might go wrong. But they’re brittle. They require someone to have anticipated the problem, translated it into logic, and hard-coded it into a system. Every rule is a bet that you already know where the risk is.
The problem is that the most expensive anomalies — the ones that actually hurt — are the ones nobody thought to write a rule for.
- The vendor that has been slowly inflating line-item prices by 2% per quarter across 47 locations.
- The GL coding pattern that shifts during a simple surge of staff turnover in ways that quietly corrupt your departmental P&L.
- The supplier who’s suddenly splitting invoices just below your approval threshold — not because of fraud, necessarily, but because someone on their side changed how they bill and nobody on yours noticed.
These are fuzzy problems. They don’t trip thresholds. They live in the space between “Everything looks fine” and “How did we miss $400K in cost creep?”
MCP Changes the Nature of the Question
What MCP does — and I mean this practically, not as a thought experiment — is give an AI agent live, structured access to your actual production data. Not a summary. Not a pre-aggregated report. The real tables with real joins are filtered in real time.
At Ottimate, I’ve been working directly with our internal production database through MCP-connected agents. And the shift in analytical capability is not incremental. It’s categorical.
Here’s a concrete example. I can sit down and ask: “Look at our invoice data for the hospitality vertical. Are there any vendors where the average line-item unit cost has increased quarter-over-quarter for three or more consecutive quarters, but total invoice volume hasn’t changed?” That’s not a canned report. There’s no dashboard widget for it. It’s a hypothesis — barely formed — that I can test in the time it takes to type the question.
The agent writes the SQL. It handles the joins. It accounts for the data quirks — our invoices table has 152 million rows with no date partitioning, so it knows to filter by account ID ranges against the sort key rather than attempting a full scan. It returns results, and then — this is the part that matters — I can have a conversation about what those results mean.
“OK, now cross-reference these vendors against accounts where we’re seeing higher-than-average processing times. Is there a correlation between price creep and operational friction?”
That second question would have taken a data team days to scope, write, validate, and present. With MCP, it’s a follow-up sentence.
Anomaly Detection Becomes Exploratory, Not Prescriptive
This is the paradigm shift that I think most people in finance haven’t fully internalized yet. Anomaly detection is moving from prescriptive rule sets to exploratory conversation.
In the old model, you needed three things to catch a problem: awareness that the problem category existed, technical skill to query for it, and time on a data team’s roadmap to build and maintain the detection. That’s a high bar. It means the vast majority of anomalies — the fuzzy, emergent, context-dependent ones — simply go undetected until they’re large enough to show up in a quarterly variance report, at which point you’re doing forensics instead of prevention.
MCP-connected agents collapse that entire chain. The finance operator becomes the analyst. The question becomes the query. And because the interaction is conversational, you can iterate in real time — narrowing, broadening, pivoting — in ways that mirror how experienced finance people actually think about problems.
The best AP managers I’ve worked with have always had a sixth sense for when something “feels off” in their data. They’d notice a pattern, squint at a report, and pull a thread that led somewhere important. The problem was that pulling that thread required either SQL fluency or a request to someone who had it. MCP removes that bottleneck. The intuition becomes directly actionable.
The Skills That Matter Are Changing
Here’s where I want to push the conversation a bit further, because I think there’s a workforce implication that’s being underestimated.
The finance professionals who thrive in this environment won’t be the ones who are best at building pivot tables or writing VLOOKUP formulas. They’ll be the best at asking questions. Specifically, the ones who can:
- Frame a vague suspicion as a testable hypothesis.
- Know enough about their data landscape to guide an agent toward the right tables and relationships.
- Interpret results with business context that no model has — the knowledge that a specific vendor renegotiated terms in Q3, or that a location manager changed, or that a seasonal pattern exists that doesn’t show up in 12 months of data.
This is a fundamentally different skill set than what finance teams have been hiring for. It’s less about technical execution and more about analytical intuition combined with domain expertise. The people who’ve been closest to the operational reality of their business — the ones who know where the bodies are buried, so to speak — suddenly have a force multiplier they’ve never had before.
But it requires adaptability. And I want to be honest about that. I’ve watched people who are extremely good at their jobs resist this shift because it feels like the ground is moving under them. It is. The answer isn’t to pretend it isn’t — it’s to recognize that the value of deep operational knowledge just went up, not down. The bottleneck was never the insight. It was the tooling and output capacity to act on it.
What This Means for Finance Leadership
If you’re running a finance function right now, here’s what I’d be thinking about:
Your anomaly detection strategy is probably too rigid. If it’s entirely rule-based, you’re catching the problems you already know about and missing the ones that actually matter. MCP-connected agents give you a way to complement those rules with exploratory analysis that can surface the unknown unknowns — not as a replacement for your existing controls, but as a layer on top of them.
Your team’s analytical capacity just got multiplied, but only if you invest in the right skills. The people who can bridge business context and data exploration — who can sit with an agent and iteratively work on a problem — are going to be disproportionately valuable. Start developing that muscle now.
And, most importantly, the organizations that figure this out first will have a structural advantage. Not because the technology itself is proprietary — MCP is an open protocol — but because the combination of clean, connected data, domain expertise, and a culture willing to explore is genuinely hard to replicate.
We’re building toward this at Ottimate. Not as a feature announcement or a roadmap bullet point, but as a fundamental rethinking of what a finance platform should enable. The goal isn’t to give you better reports. It’s to give you the ability to ask questions you couldn’t ask before — and get answers fast enough to actually do something about them. The Ottimate Copilot is here today, and it’s just the beginning.
The future of finance ops isn’t more dashboards. It’s not a prompt waiting for you to ask a question. It’s better questions and agents to help you answer them.
Want to see how Ottimate Copilot helps your team uncover hidden insights and act faster? Book a demo with our team.