Most leaders agree that AI will not fix a broken process. True. But this is exactly why SMEs should use AI as the forcing function to choose one way of working, encode it, and remove the daily dependence on tacit knowledge. Policies live in PDFs, work lives in people’s heads. New joiners copy Bob or Jenny, not the quality manual. Auditors then discover three versions of the truth. AI gives you the chance to tidy the house, once, and keep it tidy.

The real problem

It is not the absence of documents, it is the absence of a single, executable way of working. We write policies, we onboard quickly, we hope good sense fills the gaps. It does not. Juniors revert to what school taught them, seniors import habits from previous employers, and the organisation drifts. As the French say, the gap between how we say we work and how we actually work becomes, how shall we put it, généreux.

The better approach

Use AI to turn your process into a guided workflow that people cannot ignore and do not need to remember.

1. 1. Set outcomes and guardrails
      Decide the result you want, cycle time, quality bar, approval rights, data residency, audit trail.

1. 1. Map reality, not theory
      Shadow the work. Document the actual steps, the handoffs, the exceptions. No pride, just facts.

1. 1. Define the golden path
      One optimised way of working for 80 percent of cases, with named owners for the rest.

1. 1. Encode the workflow
      Build it into a workflow engine, add an LLM assistant that retrieves the right artefact at each step, attach forms, rules, and service levels, trigger checks and approvals automatically.

1. 1. Make knowledge executable
      Replace static SOPs with prompts, decision trees, and generated checklists. The system serves the next best action, the person applies judgement.

1. 1. Instrument everything
      Track lead time, error rate, rework, handoff delays, first time fix, customer effort, and show the numbers to teams weekly. Voilà, continuous improvement with less theatre.

1. 1. Train by doing
      New starters follow the workflow on day one. They learn your way, not Bob’s, not Jenny’s, and not the last firm’s.

What this looks like in an SME

Take quote to cash for a scientific instrument maker.
Incoming enquiry is classified, the workflow opens the correct path, price guidance appears with margin guardrails, required approvals route automatically, the LLM drafts the proposal from the current template and product data, the system checks export controls and credit before booking. Handoffs to engineering, purchasing, and service are triggered with clean data. When the auditor visits, the trail is already there. No hunt through mailboxes, no folklore.

Service is similar. Intake, triage, and troubleshooting trees are encoded. The assistant suggests probable faults from case history, parts lists are proposed, logistics and risk checks run in the background, and the customer receives a precise ETA. Humans handle the hard calls, the rest flows.

Measurable gains

Expect fewer errors, faster cycle times, cleaner data, and a calmer team. Typical early wins in SMEs
• 30 to 50 percent reduction in handoff delays
• 20 to 40 percent drop in rework and exceptions
• First time fix up, warranty leakage down
• Audit findings move from major to minor

Numbers vary, the principle does not, when the workflow runs the process, people can focus on value.

Common traps and how to avoid them

• Automating chaos, slow down, map reality first.
• Tool first thinking, start with outcomes and ownership.
• One big bang, deliver in two week slices, one process at a time.
• Letting the manual live on as a parallel truth, retire it, keep it for compliance, not for daily operations.

A simple 30, 60, 90 day plan

Days 1 to 30, pick one high volume process with pain, map it, define the golden path, design the data you need.
Days 31 to 60, build the workflow, connect to your core systems, add the assistant where it saves time, pilot with one team.
Days 61 to 90, tighten rules and prompts, publish dashboards, train by doing, switch off the old path, then pick the next process.

Final word

AI will not rescue bad process thinking, agreed. But used correctly, it will force clarity, codify the way you work, execute the routine steps, and keep the discipline every day without nagging. Less folklore, more flow. If that sounds refreshingly unglamorous, good, that is where the money is.