Six situations, six design hypotheses
For each situation common in SMEs: what happens today, the solution we would consider, the rules of use and the results to measure. These are design examples, not client references.
Finding the right information
Situation
In a service SME, procedures, customer histories and technical documents are spread across email, a file server and two business applications. Every search takes several minutes and the answer depends on who does it.
Proposed solution
A search assistant connected to the authorised sources, which answers with excerpts and their sources rather than an answer to take on trust. Existing access rights are respected; nothing is copied outside the company without a rule.
Rules and results to measure
The rules define the indexed sources, access and how long queries are kept. Indicators: search time, share of answers with a source, questions left unanswered.
Handling customer requests
Situation
Requests arrive by email and through a form, are read one by one, forwarded to the right person, sometimes forgotten. Answers to frequent questions are written from scratch every time.
Proposed solution
An existing AI tool categorises each request, drafts a reply from validated answers and directs it to the right person. Sending remains a human decision: nothing goes out without review.
Rules and results to measure
The rules define the data read, the categories and the requests that require a direct human answer. Indicators: time to first reply, share of correctly routed requests, drafting time.
Preparing quotations in an engineering office
Situation
Requests arrive by email with heterogeneous attachments. A project manager re-enters the elements, looks up prices and writes the quote in a spreadsheet.
Proposed solution
An existing AI extracts and structures the request; an automation matches it with the catalogue and previous quotes; a simple interface presents a proposal. The project manager checks, adjusts and decides.
Rules and results to measure
The usage framework specifies the data used and the validations. The indicators tracked are response time, preparation time and error rate.
Producing documents from existing data
Situation
Meeting notes, service reports, product sheets or letters are written by hand from information already present in the tools, with inconsistent templates.
Proposed solution
An automation gathers the data from your tools, an existing AI tool drafts a first version following your templates, and a simple interface lets you review, correct and approve before distribution.
Rules and results to measure
The rules define the templates, the data used and the mandatory review before sending. Indicators: production time, corrections at review, distribution lead time.
Removing re-entry between tools
Situation
Orders entered in the sales management tool are re-entered in the planning spreadsheet, then in the invoicing tool. Discrepancies come to light at month-end.
Proposed solution
Connections between the tools, through APIs or connectors developed for the purpose, let validated information flow once; a check flags discrepancies instead of letting them through. No AI is needed here.
Rules and results to measure
The rules describe the flows, the check rules and who corrects what. Indicators: re-entry time removed, discrepancies detected, invoicing lead time.
Keeping deadlines and service quality
Situation
Files sit waiting without anyone noticing. The customer follows up, and that is how the delay comes to light.
Proposed solution
Automatic tracking of deadlines on open files, an alert to the right person before the due date and a simple dashboard. An existing AI tool can summarise the state of a file; the decision remains human.
Rules and results to measure
The rules define the thresholds, the alert recipients and the tracking data. Indicators: overdue files, average handling time, customer follow-ups.
Let’s talk about your priorities
Do you recognise one of these situations? A first conversation of 30 minutes to an hour, by video call, to see whether it deserves a diagnosis.