Fifteen things AIcan do today.
No brochure futurology. Fifteen concrete applications I build today inside your own Microsoft 365 environment — with a person checking the result.
All examples are representative and anonymised.
Fifteen applications. All buildable.
From a single concrete step to a complete process. Filter by what you recognise.
- 01
Invoices that read themselves
The invoice arrives. The details are already in.
How it worksInvoice hits the mailbox → AI fills in vendor, amount and date → someone approves or declines.
What disappearsRetyping PDFs into Excel or your bookkeeping.
Documents - 02
Meetings that take their own notes
You talk. The minutes appear.
How it worksRecording of a client call → AI transcribes and extracts actions and decisions → the summary lands in the file.
What disappearsRe-listening and summarising yourself.
Office - 03
Handwritten notes, simply readable
A scribble on paper becomes a clean record.
How it worksPhoto of a filled-in checklist or receipt → AI reads the handwriting → the details land in the app.
What disappearsNotes that stay on someone's desk.
Documents - 04
Documents that compare themselves
What changed between version 3 and 12?
How it worksTwo contract versions in → AI marks every change → the reviewer only looks at the differences.
What disappearsComparing line by line.
Documents - 05
Ask a question, get an answer with sources
“Where was that again?” — asked of your own files.
How it worksEmployee asks in Teams → AI searches procedures, files and notes → answer with a link to the source.
What disappearsInterrupting colleagues, or getting lost in folders.
Knowledge - 06
Email that sorts itself
Every incoming mail knows where it belongs.
How it worksClient mail arrives → AI recognises the request type → task to the right colleague, summary on top.
What disappearsPloughing through the shared inbox together.
Office - 07
Processes that move themselves along
AI assesses within the rules and carries on.
How it worksRequest meets all agreements → handled automatically → only exceptions reach a human.
What disappearsThe queue of in-between jobs.
Autonomous - 08
Documents that start themselves
The first draft is ready. You make it yours.
How it worksQuote request arrives → AI drafts from your previous quotes → you refine and send.
What disappearsStaring at a blank page.
Documents - 09
Client questions answered instantly
Outside office hours too.
How it worksFAQ on the site → AI answers from your own knowledge → complex questions go politely to a human.
What disappearsTyping the same reply again.
Clients - 10
Files that summarise themselves
Twenty emails. One paragraph.
How it worksFile of forty documents → AI summary with players, deadlines and status → a new colleague is up to speed in five minutes.
What disappearsWorking through the whole file yourself.
Knowledge - 11
Approvals with a pre-check
AI flags what needs attention. The rest is routine.
How it worksRequests within the rules flow through; deviations get a flag with an explanation.
What disappearsEvery request past a manager.
Office - 12
Data that checks itself
Duplicate relations, missing fields: AI taps you on the shoulder.
How it worksNew relation entered → AI recognises it already existed → merged instead of duplicated.
What disappearsCleaning up when it's already too late.
Autonomous - 13
Planning that thinks along
Who can take this job, given capacity and preferences?
How it worksNew assignment → AI proposes two or three realistic allocations → the planner picks.
What disappearsShuffling hours in a spreadsheet.
Autonomous - 14
Messages in the right tone
Plain professional language, not consultant-speak.
How it worksShort note → AI turns it into a clear client email → check, send, done.
What disappearsTwenty minutes hunting for the right sentence.
Clients - 15
From photo to inspection report
A photo on site, a complete report back.
How it worksEngineer photographs the damage → AI recognises what's in the shot → the inspection report fills itself.
What disappearsWriting the report in the evening.
Documents
This is what AI looks like in a real working environment.
Views from solutions I have built. Rebuilt and anonymised: no client names, no client data, no company branding.
The document clearly shows an expiry date of 14 March 2032, which differs from the current value.
ApplyThe passport was issued by a non-EU country, so this should be updated to 'No'.
ApplyThe nationality in the document matches the current value.
MatchesAI compares the document with the data on file
The client asks for a short presentation of this year's key technology developments, to be used in a webinar. Based on the email exchange, the team gathers information on the tools and automations in use.
A proposal that prefills the create form. Everything stays editable.
AI suggests where an email belongs. The employee accepts, edits or dismisses.
From email to storage, with a review moment in between.
What emerges when the separate AI steps are tied together.
The client fills in the details; AI checks the application before submission.
Shipments from non-EU countries fall directly under the new charges.
265 shipmentsModelling new rules against the company's own shipment data.
Raw transaction data underpinning the analyses, with status per line.
A reduced rate applies to food intended for human consumption. A personalised finish does not change the nature of the product.
If the non-food element is significant, its treatment should be assessed separately.
AI determines the rate from HS code and description, with the source attached.
Representative reconstructions of functionality that was built. AI proposes; a person decides.
One of the fifteen — or something not on the list yet.
Tell me where it hurts in your business. I'll think along about what AI can realistically do there, then build it myself.
No strings attached — you'll hear from me personally within one working day.