Practical AI in Modeling: Four Skills You Can Implement Today
Oct 26, 2026Finance teams are gaining access to increasingly powerful AI tools, yet the underlying work does not always improve with the technology. Reports may still depend on manual exports, copied formulas, ageing macros, and workarounds understood only by the people who created them. Introducing AI into this environment may accelerate individual tasks while leaving the wider process untouched.
This challenge shaped dbrownconsulting’s webinar, Practical AI in Modeling: Skills You Can Implement Today, hosted by David Brown with Giles Male, Co-founder of Full Stack Modellers and widely known as the Humble MVP.
Their conversation focused on four practical capabilities finance professionals can develop to apply AI more deliberately: evaluating existing workflows, translating professional standards into reusable instructions, distinguishing fixed rules from uncertain assumptions, and assigning ownership to AI-enabled work.
These skills begin with the work itself, because choosing an AI tool before understanding the process can result in sophisticated technology being applied to the wrong problem.
- Audit the Work Before Introducing AI
Start by listing the recurring activities in your role and assessing each one according to the time it consumes and the complexity or value of the problem it solves. David’s WorkIQ framework places these activities into four categories:
- Eliminate busy work that consumes time without creating sufficient value.
- Automate clear and repeatable tasks governed by stable rules.
- Improve complex processes weakened by poor design or capability gaps.
- Protect high-value work that depends on expert judgment.
This assessment prevents AI from becoming the automatic answer to every productivity problem. A workflow built around exporting data, copying it into Excel, creating a pivot table, linking cells, adding a VLOOKUP, and running a macro may appear ready for Copilot. However, Power Query training or a redesigned reporting process may solve the problem more reliably.
Once the right task has been identified, the next challenge is consistency. AI needs more than a request. It needs access to the standards by which good work is judged.
- Turn Professional Standards Into Reusable AI Skills
Correcting the same formatting, structural, and modeling errors in every AI conversation wastes the time automation should save. AI Skills provide a way to document those expectations once and apply them repeatedly.
During the webinar demonstration, Giles showed how personalised instructions, workbook rules, and Markdown-based Skills could guide Copilot in Excel. A financial modeling Skill might require the tool to avoid merged cells, use structured tables, apply approved colours, format dates consistently, and follow the organisation’s conventions for imports, exports, and time-series data.
These instructions turn personal preferences and house styles into reusable operating standards. However, they still require review. AI may follow a detailed rule correctly in one workbook and deviate from it in another. Skills improve consistency, but professional oversight remains responsible for quality.
That oversight becomes especially important when a model contains both fixed calculations and assumptions about an uncertain future.
- Separate Fixed Rules From Areas of Uncertainty
Financial models combine deterministic logic with probabilistic judgment. The distinction determines where AI should follow instructions exactly and where it may explore several possible outcomes.
A terminal-value calculation, tax computation, or three-statement model structure should follow an approved methodology consistently. Its logic should not change because the AI selected a different approach during a new conversation. Uncertainty belongs in assumptions such as oil prices, exchange rates, demand, or other variables that genuinely require scenarios and probability ranges.
As Giles observed, a valuation model should not “roll a dice” when deciding how to calculate terminal value. The calculation must remain dependable even when the assumptions change. Clear rules protect the integrity of the model, but rules alone cannot govern an AI-enabled workflow. Someone must also remain responsible for the system and its consequences.
- Give Every AI Workflow an Accountable Owner
David illustrated this governance problem through the story of Agent Penny, an internal digital worker that performed well until something went wrong. When the organisation attempted to resolve the failure, nobody could identify who owned the agent or had authority to correct it.
Every AI workflow therefore needs clear answers to five questions:
- Who owns and maintains it?
- Which data can it access?
- What decisions may it make?
- When must it escalate to a person?
- Who validates the final output?
Accountability cannot sit with the technology. Finance teams handle sensitive information and support decisions with material consequences, so human ownership must remain visible from implementation through review.
Start With One Controlled Workflow
Giles’s closing advice was practical: begin with a small, isolated task rather than handing an entire month-end process to AI. Test the result, document the limitations, correct the workflow, and expand only when the system proves reliable.
dbrownconsulting’s bespoke AI for Work, AI for Strategic Leaders, and AI for Board Members programmes help professionals and organisations apply this discipline to their actual roles, processes, data risks, and strategic priorities.
Practical AI implementation begins with four connected capabilities: understand the work, document the standard, protect the logic, and assign accountability. Once those foundations are in place, AI becomes easier to use and safer to scale.
