The role nobody talks about, but every finance deployment quietly depends on
A new role is emerging inside finance teams and the vendors that serve them. No stable title yet. Some call it the accounting engineer. Others, the finance person who codes. At Consark, we call it the Forward Deployed Accountant, or FDA.
Whatever you call it, this person is now the single biggest variable in whether an AI finance deployment lands or stalls.
This piece is about why.
What an FDA really is
An FDA is a hybrid. Not in the soft sense of knowing a little of both. In the operating sense of sitting on either side of the table without translation loss.
On one side, an accountant. A CPA, a senior accountant, a controller who has lived inside the close. They know how a prepaid behaves in month seven of a lease modification. They know why intercompany never clears on the first pass. They know what the auditor will ask before the auditor knows.
On the other side, a builder. Not vibe code. They write production code, design data flows, configure agents, work with APIs, and know which architectures will survive a Big 4 review.
They are not engineers who learned a bit of accounting. They are professional accountants who code, and they treat that combination as their craft.
Put them in a room with a Controller and they speak fluent close. Put them in a room with a backend engineer and they speak fluent system. The gap that normally eats months of every implementation collapses, because in the FDA’s head, the gap was never there.
Where the role comes from
Palantir pioneered the Forward Deployed Engineer over a decade ago. Instead of selling software and handing it to a systems integrator, you embed your best engineers inside the customer. They write production code in the customer’s environment. They own the outcome. They stop when the customer is getting value, not when the contract is signed.
Anthropic and OpenAI are both copying it now. AI is not shrink wrap. Every customer’s data, workflow, and risk tolerance is different. Someone must be embedded.
Now apply that to finance.
The close is one of the most context heavy processes in any enterprise. Every chart of accounts is different. Every materiality threshold is different. Every revenue cutoff is its own animal. A generic AI tool dropped into that environment is wrong on day one.
Send a software engineer and they will get the plumbing working but not know what the agent should actually do. Send a traditional accounting consultant and they will know what the agent should do but cannot configure it. Both paths leak time and quality.
The FDA does not.
Why this role is necessary now
Three things have changed at once.
Finance software has gone agentic. Not workflow tools that automate clicks. Agents that prepare accruals, run reconciliations, do flux, and validate revenue continuously. Configuration in this world is closer to programming than to administration.
Customers are no longer patient. The era of two-year ERP style implementations is over. Finance leaders have seen what good agents can do and they want time to value in weeks, not quarters.
And the talent finally exists. Ten years ago, finding a controller who could write Python and read an architecture diagram was rare to the point of mythical. Today, still uncommon, but no longer mythical. A generation of finance professionals grew up alongside the tooling. They got tired of waiting for IT. They wanted to automate the worst parts of their own jobs. The line between finance and data got too thin to ignore.
They exist. They are the pool we draw from.
What an FDA does inside a Consark deployment
This is where the abstract concept becomes concrete.
When a customer signs with us, the FDA is on the call before the kickoff. They review the customer’s chart of accounts, ERP setup, close calendar, intercompany structure, reconciliation policies, and the way the team currently handles flux. They do this as an accountant would, not as a project manager checking boxes.
Then they go to the platform. They configure Noa AI Agents, our execution layer, to match what they have learned. The accruals agent is shaped to the customer’s actual accrual policy, not a template. The reconciliations agent is set up against the real bank and subledger structure. The flux agent is calibrated to the materiality thresholds the Controller cares about. The intercompany agent is mapped against the real entity structure with its real elimination rules. This is configuration work, but it is configuration that requires accounting judgement on every decision.
Where standard configuration is not enough, the FDA writes the extension. They build the integration logic, the custom validation rule, the agent prompt that captures a specific company policy. They work alongside our core engineering team when something needs to land in the product, but they do not wait when they can build it themselves inside the customer environment.
Then they stay. Not as a support contact, but as the person who watches the first close, the second close, the third close. They see what the agents got right and what they did not. They tune. They explain what the agent did to a skeptical Big 4 auditor. They translate the customer’s edge cases back into our roadmap.
Our multi agent system runs on top of a finance data fabric. That data fabric is what lets the agents work across the messy reality of an enterprise ledger without breaking. But a data fabric without a human who can shape it to a specific business is just plumbing. The FDA is the person who makes the plumbing useful.
This is what fast deployment actually means when you look behind the marketing line. It does not happen because the product is magic. It happens because the FDA is on the ground compressing what would otherwise be months of back and forth between consultants, developers, and the customer’s finance team into a single embedded person doing the work in real time.
Coordination is not execution
There is a phrase we use internally that applies just as much to deployment as it does to the close itself. Coordination is not execution.
Most implementation models in this category are coordination models. A project manager schedules the meetings. A solutions consultant writes the requirements. A developer builds against the requirements. A trainer trains the end users. Each handoff is a place where context is lost and time leaks out. The customer ends up coordinating across half a dozen vendor roles, none of whom can finish the job alone.
The FDA model collapses that. One person, embedded, who can finish the job. The customer’s Controller has one phone number, and the person on the other end of that number can both understand the accounting problem and fix it in the system the same afternoon.
This is what global finance teams at WPP, Cyient, and Yahoo have experienced with us. The scale and complexity of those environments do not get crossed by a coordination model. They get crossed by people who can move fluently between accounting judgement and system configuration without losing a step.
We take this so seriously that it shapes how we hire. At Consark, for every two pure engineering hires we add to the core product team, we add one Forward Deployed Accountant. That ratio is deliberate. It is the structural answer to a structural problem. Most software companies would call that an expensive choice. We call it the only honest way to ship enterprise finance AI.
Who actually becomes an FDA
The honest answer is that it is a small pool, but it is a growing one.
The strongest profiles we have seen come from a few directions. Practicing finance controllers who got tired of broken tools and started building their own. Big 4 audit and advisory professionals who moved into systems work and never went back. Finance leaders inside fast moving companies who picked up Python or SQL because the data team was always backlogged. CPAs from analytics and consulting backgrounds who specifically chose to learn engineering rather than treat it as someone else’s job.
What unites them is not the credential mix, which varies. It is a particular disposition. They are uncomfortable with the answer “the system cannot do that.” They treat the configuration of a finance system as a real engineering problem with real consequences if they get it wrong. They are also unwilling to outsource the accounting judgement to a developer who has never closed a set of books. That combination of standards is the FDA personality.
If you are in finance and you can already see this in yourself, you are probably already an FDA in some form. You may just not have the title yet.
What this means for the next wave of finance AI
The pattern is becoming clear. Finance is going to be reimagined by agentic systems. That much is no longer in dispute. The question that matters is which vendors and which customers come out of this transition with working systems, and which ones come out of it with shelfware.
Our bet is that the answer turns on the FDA. Vendors who have built their delivery model around embedded accountants who code will move fast and earn customer trust. Vendors who treat implementation as a project management problem will struggle to keep up, because every customer environment is different and templates do not survive contact with a real close.
On the customer side, the finance teams who win will be the ones who recognize the FDA archetype on both sides of the table. They will hire for it internally where they can. They will demand it from their vendors where they cannot.
This is why Consark.ai was built by professional accountants who code, and why we keep investing in the FDA function as the platform scales. It is not a service line bolted onto a product. It is the model.
If you are a controller, a senior accountant, or a finance professional who has been quietly teaching yourself to build, you are who we are looking for. The Forward Deployed Accountant role is permission to do the thing you already do, with a team that takes the combination seriously.
The next wave of finance is not going to be delivered by software alone, and it is not going to be delivered by consultants alone. It is going to be delivered by the people who refused to choose.
