Designing a simple decision system so AI can really help your managers

You run an SME and, despite all your tools, you often feel that rules change depending on who is in charge: a salesperson grants a discount that was never approved, a team member applies an old procedure, a manager does things “like we’ve always done it”. The result: inconsistent decisions, internal friction, unhappy clients… and a lot of time spent fixing problems. AI could help, but as long as your business rules live only in people’s heads, it can’t do much.

In this article, we’ll look at how to move from “case-by-case management” to a clear, shared decision system where AI supports your managers without replacing them. The goal: more consistent decisions, less time wasted arbitrating the same issues, and a calmer organisation.

Why your business rules stay fuzzy (and what it costs you)

In many SMEs, business rules are:

• Verbal: “We’ve always done it this way”
• Scattered: a bit in a spreadsheet, a bit in an email, a bit in an old contract
• Personal: everyone has their own way of handling exceptions

This leads to very concrete consequences:

• Inconsistent decisions: two similar customers get different answers
• Dependence on a few key people: when they’re away, everything slows down
• High mental load: managers “recalculate” the same decisions again and again
• Blocker for automation: as long as rules aren’t explicit, you can’t safely delegate part of the work to AI or simple automations

As long as your business rules are not written in plain language, AI will remain a gadget: it can write nicely, but it doesn’t know how you actually decide.

The goal is not to make everything rigid, but to separate what can be standardised (most cases) from what should remain human judgement (sensitive or exceptional situations).

Build a simple decision system before adding AI

Before talking tools, you need to design a decision system:

1. Clear rules for common cases
2. Guardrails for sensitive situations
3. An escalation path when rules are not enough

Step 1 – Pick one specific decision domain

Don’t start with “all decisions in the company”. Choose one domain, such as:

• Discount approvals
• Payment terms
• Late payment handling
• Time off and remote work approvals
• Approving or rejecting specific client requests

Selection criteria:

• The situation occurs frequently
• It creates tension or confusion today
• It has a visible impact on revenue, cash or team climate

Step 2 – Map the recurring decisions

For 1–2 weeks, ask the relevant managers to log the decisions they take in this domain:

• Context (client, amount, type of request, etc.)
• Decision taken
• Why (one sentence)

You can do this in a simple table:

• Columns: Date, Context, Decision, Reason, Complex? (Y/N)

After this short period, you will see patterns emerging, such as:

• “We usually accept up to X euros”
• “We always refuse in that situation”
• “Above this amount, the CEO must approve”

Step 3 – Turn experience into simple “if… then…” rules

The goal is not to write a legal manual, but operational rules that anyone can use – including an AI assistant.

Examples:

• If discount ≤ 5% and customer has been active for > 12 months, then salesperson can decide alone.
• If discount between 5% and 10%, then sales director must approve.
• If late payment ≤ 15 days and first incident, then send standard reminder email.
• If late payment > 30 days or recurring, then accounting manager calls the customer.

Add some explicit grey zones as well:

• If customer is strategic (predefined list), then decision is taken case by case by the CEO.

Step 4 – Design the decision flow

Once your rules are written, you can draw a small decision flow. For example, for handling special customer requests:

This kind of diagram helps your teams know when they decide and when to escalate. It’s also an ideal base to connect an AI assistant later.

Add AI as a decision assistant (without giving it the keys)

Once this simple system is in place, AI becomes genuinely useful. It doesn’t “decide” on its own, but it can:

• Gather the right information before the decision
• Apply standard rules on simple cases
• Flag exceptions that must be handled by a human

Three concrete roles for AI in your decisions

1. Case preparer

   • Summarises key elements: customer history, amounts, past incidents
   • Formats information into a standard template

2. Rule checker

   • Compares the situation with your “if… then…” rules
   • Suggests a standard recommendation (for the human to accept or adjust)

3. Exception guard

   • Alerts when escalation criteria are met (high amount, strategic customer, risk…)
   • Logs out-of-frame decisions to improve your rules over time

Concrete example: approving a sales discount

Without AI:

• The salesperson emails the director: “Customer X is asking for 12% discount, what do we do?”
• The director searches for previous quotes, looks at the account, hesitates, and decides on a case-by-case basis.

With a decision system + AI:

1. The salesperson fills a short form (customer, amount, discount requested, reason).
2. The AI assistant:
   • Retrieves customer history
   • Applies discount rules: thresholds, customer tenure, strategic status
   • Suggests a recommendation and signals if escalation is required
3. The manager:
   • Reviews the proposal
   • Confirms, adjusts or rejects it

Analysis time goes down, decision consistency goes up, and you stay in control.

How to get started in 10 days

You can set up your first AI-assisted decision system without a big IT project.

Actionable checklist

1. Day 1–2: choose your pilot domain

   • One single family of decisions
   • Clear pain today and visible impact

2. Day 3–7: log real decisions

   • Shared table
   • Each manager records decisions and reasons

3. Day 8–10: formalise and test

   • Turn recurring patterns into “if… then…” rules
   • Draw a simple flowchart
   • Configure a basic AI assistant (in a tool you already use: office suite, CRM, ticketing tool…) to:
     • Generate a standard decision sheet
     • Remind the written rules
     • Suggest a recommendation

After this pilot, improve it every two weeks:

• Capture new edge cases
• Simplify overcomplicated rules
• Clarify grey areas with your teams

Conclusion

By structuring your decisions before talking about tools, you create a clear framework in which AI can finally become a real lever instead of a shiny gadget:

• Your business rules move from people’s heads to a shared, usable format
• Managers save time on repetitive, standard cases
• Sensitive situations remain fully human decisions, but with better preparation
• AI plays its role as preparation and checking assistant, not as a black-box decision maker

If you want to go further, start with one decision domain and measure the benefits over a few weeks. You’ll quickly see where AI truly helps you – and where it should stay in the background.

If you would like support in your digital transformation, Lyten Agency can help you identify and automate your key processes. Contact us for a free audit.