- Who it's for
- Owners and managers who want to introduce AI without losing the trust of their team.
- What you'll be able to do
- Start with the most annoying routine tasks using the 30-day plan, and set the rules, approvals and training around them.
- As of
- February 2026
AI doesn’t replace people – above all it replaces manual, repetitive tasks. Introduce it as a “friction killer” (with approvals, rules and training) and you lift output per person – without shrinking the team.
Why the fear of AI is so strong (and why it often points the wrong way)
A lot of people feel the pressure right now: a weaker economy, higher costs, less appetite for hiring. Add AI that “can do everything” and the conclusion seems obvious: jobs will go.
Reality usually looks different – and it’s an opportunity. AI is extremely good at searching, sorting, summarising, drafting, classifying and evaluating. And companies lose time on exactly those things every day: email chaos, hunting for documents, reports stitched together by hand, keeping data tidy.
That makes AI less of a job killer than a friction killer. Introduce it properly and work gets measurably lighter – and people measurably more effective.
The most important shift in perspective: jobs vs tasks
When people talk about “jobs”, they usually mean a role. In practice, though, it’s bundles of tasks that get automated first.
The AI takes on routine research, first drafts, standard replies, extracting data and plausibility checks. What stays with people: the decisions, the customer conversations, risk and accountability, quality, prioritisation and context.
That’s the good news: roles are less likely to disappear completely – they move up.
Empower instead of cut: 5 principles that work in companies
- 01Automate the most annoying thing firstTasks that happen often and carry little risk. The result: relief you feel immediately, and less scepticism.
- 02Human approvalThe AI makes suggestions and drafts, people approve — and anything uncertain goes to a human.
- 03Make quality measurableTime per case before and after, turnaround from enquiry to quote, correction effort, error rate.
- 04Guard rails instead of gut feelingA one-page policy: which data may go in, what is off limits, which sources count, what gets logged.
- 05Career paths for your peopleTraining per team, roles cut differently and clear responsibilities — the AI suggests, you decide.
1) Automate the most annoying thing first (not the most critical)
Start with tasks that happen often and carry little risk: pre-sorting emails, meaning prioritising, summarising and drafting replies. Documentation and minutes, turning notes into a clean summary. Reports that pull several sources onto one page. And filing, so that folders and naming stay tidy.
The result: relief you feel immediately, and less scepticism.
2) Human approval – but consistently
“The AI decides” creates fear. A clear split works better: the AI makes suggestions and drafts, people approve — and when in doubt, one fixed rule applies: anything uncertain goes to a human.
3) Make quality measurable
Without measurement you’re guessing. Four figures are enough and easy to collect: the time per case before and after, the turnaround from enquiry to quote, the correction effort — meaning how much of the draft survives — and the error rate including complaints.
4) Guard rails instead of gut feeling (policy + data)
A one-page policy is often enough. It answers four questions: which data may go into AI workflows? What is off limits? Which sources count? And what gets logged, who approves?
5) Career paths for your people (instead of automating them away)
Say openly that tasks are changing. Then deliver the plan that goes with it: 60 to 90 minutes of training per team, roles cut differently with less admin, more customer work and more process quality, and clear responsibilities — the AI suggests, you decide.
Practical examples (small / medium / large)
A) Trades business / local service provider (10–30 employees)
Typical friction: enquiries slip through, quotes take too long, follow-ups happen irregularly, documentation and filing eat time.
Example workflow (close to Google Workspace, but connectable via API):
-
Enquiry assistant Gmail/form → AI structures it (customer, location, service, urgency, photos) → creates a ticket or task.
-
Quote draft AI creates the first draft from building blocks → a human adjusts it and sends it.
-
Automatic follow-up No answer after X days → a follow-up draft lands in Gmail.
-
Documentation from voice notes Speak briefly into your phone → AI produces clean documentation and files it in Drive.
Effect: faster responses, less chaos, people spend more time with customers instead of in the inbox.
B) Mid-sized company (100–800 employees)
Typical friction: knowledge is scattered, documents can’t be found, processes are slow, reports cost nerves.
A knowledge system with sources searches Drive, Confluence and shared folders and answers with citations, so everyone can see where something comes from. Document automations read fields out of PDFs, check them for plausibility and hand them over — after approval — to the connected systems. And a status report for management pulls ERP, CRM, customer service and web analytics onto one page: what happened? What is critical? What should we do?
Effect: less rework, faster decisions, fewer knowledge silos.
C) Knowledge-heavy teams (agency/software/consulting)
Typical friction: briefs are unclear, context gets lost, quality fluctuates.
Notes from a call turn into a brief with tasks and open questions. A research session turns into grouped findings, a summary and a source list. And before anything is approved, a quality check runs that uses checklists against invented details and empty phrases.
Effect: more work finished per person, better consistency, less stress.
The 30-day plan: introducing AI without destroying trust
If you take away only one thing: start small, deliver measurable value, then scale.
Week 1: choice + rules
You pick one use case — frequent, annoying, manageable risk —, define what goes in and what should come out, and write the policy on a single page: data, approvals, logging.
Weeks 2–3: build the first version (with human approval)
The first version doesn’t have to be pretty, it has to run. Test it on 20 to 50 real cases, anonymised if need be, and measure time, quality and acceptance as you go.
Week 4: stabilise + next use case
Now you catch up on the edge cases, train the team for 60 to 90 minutes and take on the second use case from your list.
- 1 clear use case (frequent + annoying)
- input data and result defined (including what’s off limits)
- approval flow (who reviews what?)
- logging/transparency (audit trail)
- metrics (time, errors, turnaround)
- training + responsibilities
A positive AI future: what that means in practice
An AI-positive future isn’t “everything gets automated”. It’s more productivity per person, so companies stay competitive. It’s better work, because the admin swamp shrinks. It’s training as standard, because roles move up instead of disappearing. And it’s more speed in delivery, because tools and automations lower the hurdles.
AI quick start with Triple A Digital
I don’t build “AI showcase projects”, I build workflows that measurably take load off teams – with safeguards, approvals and quality assurance.
What you get:
- 10–20 concrete use cases for your company
- Prioritisation by value and feasibility
- Data and tool check (e.g. Google Workspace, CRM, accounting, internal systems)
- A roadmap for 30/60/90 days and a work plan for the first use case in production
Send me: your team size and the top 3 processes that are eating time right now – I’ll suggest the best use cases for the quick start.