Expense submissions are one of the most common points of friction between employees and finance teams. Employees submit reports without knowing whether something is reimbursable, finance sends them back for missing documentation, and the cycle repeats. It is not a motivation problem. It is an information access problem.
A context-engineered Expense Policy Enforcer gives employees the guidance they need before they submit, not after. Instead of guessing whether a client dinner qualifies or whether a receipt is required for a $45 purchase, they can get a clear answer at the point of decision.
RAG #6 in the Kendall Project's 100 Essential Enterprise AI Assistants is the Expense Policy Enforcer. This post walks through what one looks like and what you need to build it.
What an Expense Policy Enforcer Does
The Expense Policy Enforcer is a context-engineered AI assistant built on a RAG model. It gives employees accurate, role-specific guidance on expense submission before they file, reducing rejection rates and the back-and-forth that follows.
At its most useful, it handles questions like:
- Is a client dinner for four people reimbursable, and what is the per-person limit?
- What documentation do I need for a hotel stay over three nights?
- My manager is traveling with me. Who submits the team meal expense?
- Can I book business class for a flight over eight hours, and do I need pre-approval?
- What is the difference between the entertainment limit and the internal team meal limit?
- My expense is over the approval threshold. Who needs to sign off and how do I route it?
The difference between a generic AI assistant and a context-engineered one is significant here. A general-purpose AI can describe what expense policies typically look like. An Expense Policy Enforcer built on your organization's context can tell a sales manager in your Chicago office exactly what the client entertainment cap is for their region, whether a particular vendor category requires pre-approval, and which form to attach for international travel reimbursement.
The Context Blocks That Make It Work
Rules: The Rules block is the core of the assistant. It captures what is reimbursable, what is not, the limits that apply by expense category, the documentation thresholds, and the exception processes. The most important structural work is capturing the conditional logic: limits vary by expense type, purpose, and role, and that structure needs to be explicit so the assistant can reason through it rather than retrieve a flat answer. Pre-approval requirements belong here too.
Processes: The Processes block covers how to submit, which system to use, how to attach receipts, how to route for approval, and how to handle exceptions. Approval routing logic is worth particular attention. If the Processes block captures it clearly, the assistant can tell an employee upfront exactly whose approval they need and how to request it.
Roles: Expense policy is rarely uniform across roles. Approval thresholds differ by level. Travel class eligibility may differ by seniority or trip purpose. The Roles block maps those distinctions so the assistant gives role-appropriate answers. Manager-specific rules also belong here.
Locations: Expense policy often varies by country or region. Per diem rates differ by city. VAT reclaim procedures apply in some jurisdictions and not others. The Locations block captures those variations. For organizations with international operations, this is often where the most complexity sits and where the most value comes from.
Glossary: Expense policies use terminology employees frequently misinterpret. What counts as "client entertainment" versus a "business meal"? What qualifies as a "lodging incidental"? The Glossary block defines those terms precisely so the assistant uses them consistently and guides employees to the right category rather than leaving it to interpretation.
Where to Start
The most productive first step is mapping the Rules block with a focus on the conditional logic rather than flat policy statements. Most organizations have expense policy documented somewhere, but it is often written for a human reader who can infer context. Restructuring it so the conditions, limits, and categories are explicit is the work that makes the assistant accurate.
The Locations block is typically where the most effort is required for organizations with international operations. Regional rates, currency policies, and country-specific procedures often exist in scattered documents or are known primarily by the finance team members who handle those regions. Surfacing and structuring that content pays off well beyond the AI project.
The Expense Policy Enforcer is a way to make accurate policy guidance available at the moment an employee needs it. Getting the context right is what makes the difference between an assistant that gives employees a policy document to read and one that tells them exactly what applies to their situation before they submit.
This is Blog 6 of 100 in the series. Next Friday: Procurement Navigator (#7).