He does not want AI to do the CFO’s job. He wants it to get out of the way.
We met Hunter Bergschneider at a small dinner in New York City. He is the founder and principal of Mountain Cutter, a strategic finance and operations consultancy for early-stage startups, scale-ups, and their investors. He tried NumbersGame because he needed current QuickBooks Online books inside the work already happening: the model, the board deck, and the client call.
Build the finance function before there is a finance team.
Mountain Cutter provides fractional CFO leadership and finance-function support for startups and investors. The work spans fundraising, budgets, financial models, operating systems, accounting oversight, board management, and the decisions between them.
- Fundraising readiness: help founders and investors work from current financial information when raising capital.
- Planning and operating cadence: build budgets, models, and finance systems that make decisions easier to review.
- Accounting oversight: identify what needs attention, give accountants precise instructions, and retain a responsible accountant for the numbers.
Mountain Cutter describes itself as a strategic finance and operations consultancy for early-stage startups and scale-ups. Learn more at mountaincutter.com.
The numbers were in QBO. The work was everywhere else.
Hunter manages three to four clients at once, almost all on QuickBooks Online. For a client with four entities, getting data into his models and tools meant downloading, uploading, and copying between systems. The spreadsheet assembly could take more attention than the CFO work it was meant to support.
Too much retrieval
Before NumbersGame, moving current numbers from multiple QBO files into a working model meant repeated exports and manual handling.
Too little context
Financial data needed to be evaluated alongside models, emails, Slack conversations, call transcripts, and board decks, not as an isolated download.
A decision-making delay
On founder and accountant calls, Hunter needed real-time financial context to ask better questions and move the conversation forward.
He did not need another platform. He needed to stop copying.
For Hunter, a broad new finance stack could easily become overkill. NumbersGame offered a more useful first test: connect the books, bring them into the conversation, and see whether the actual work moved faster.
The answer was not a fantasy of replacing the finance function. It was a concrete starting point: current books in the same environment as the materials a CFO already uses to prepare, challenge, and explain a decision. As frontier models improved, that operating context became more valuable, not less.
An AI-native CFO has the data in the conversation.
NumbersGame gives me real-time data and insights to support client calls. With the books in the same environment as my financial models, emails, Slack, call transcripts, and board decks, my LLMs can identify what needs attention and help me give accountants precise, actionable instructions to finish the work correctly.
The data is not the decision. It is how he gets to one.
For Hunter, the value is NumbersGame’s multi-connect workflow combined with its GAAP accounting skill. It puts approved books in ChatGPT and Claude with the surrounding context a serious review needs, without pretending the model gets the final word.
Connect approved QBO files
NumbersGame’s multi-connect workflow lets Hunter access approved client books without a download-and-copy routine for each question.
Review the books in the working context
Within ChatGPT and Claude, Hunter can put current ledger data beside the materials that explain the business: forecasts, board packs, correspondence, and call context.
Prepare the next conversation
He can identify issues, test how a revised financial statement could affect a model or board deck, and give the accounting team a clear, actionable follow-up.
Consolidate and investigate
When a client’s structure calls for it, Hunter can bring connected-entity financials into the review and use frontier models to surface industry-specific questions, including healthcare considerations, for qualified follow-up.
Keep the separation of duties
Hunter can perform the analysis, but he keeps separation of duties and a proper accountant responsible for the final numbers and accounting changes.
He is building MCOS because the model is not the asset.
Hunter works across ChatGPT, Claude, and Gemini, including voice-led workflows while tasks advance in the background. It reminds him of the private-equity rhythm: give marching instructions, then come back when the model is ready. The difference is that the work can now move while the conversation is still happening.
- MCOS: Mountain Cutter’s context, workflows, standards, macros, and quality checks that can carry across models.
- Quality is the work: model drift and hallucinations can turn simple work into rework. Hunter is building the continuity and repeatable quality needed to make AI useful across a practice.
- Scale delivery, not just speed: reliable AI makes individual tasks faster, but Hunter’s aim is capacity for more useful work, not merely fewer hours worked.
- More capacity for CFO judgment: use AI for preparation and follow-through, then spend the saved capacity understanding the business, challenging assumptions, and working through decisions with founders and investors.
Move fast with AI. Start with client consent.
Hunter’s advice to other CFOs is direct: be open with clients about AI processing, make third-party AI processing explicit in the engagement contract, and understand any HIPAA or other compliance requirements before connecting client data.
He starts with work he knows well, proves what is useful, and keeps building from there. That is the practical AI operating model for a fractional CFO: real client work, explicit permissions, reviewable outputs, and professional responsibility where it belongs.
The CFO role is changing. Hunter is learning in public.
Hunter sees AI fluency as a professional imperative for CFOs and accountants. His approach is not to treat a model as the finance function. It is to learn fast enough to pair stronger tools with stronger judgment, so the practice can keep delivering the strategic work founders and investors need.
See how a fractional CFO works with live financial context.
Bring a QuickBooks Online question, a model, or a multi-entity reporting problem to a NumbersGame walkthrough.