Knowledge
Knowledge
Experts with the right tools, and what we learn along the way.
We set up and manage AI workflows for SMEs. Here we share the frameworks we use: which processes are suitable, where people decide and what management requires. Your AI grows with your organisation: you keep moving, we manage it.
Decisions and control
What does an AI agent do in a business process?
The boundary between proposing and executing, made concrete. What an agent does, and what you leave for a person to decide.
Email, Documents
Google Workspace or Microsoft 365 as a foundation
What your stack means for route, permissions and admin load. An honest comparison, including where the connection runs differently today.
Decisions and control
The second implementation as proof
Once is a demo, twice is a pattern. What the second implementation demonstrates and what you need to measure.
Decisions and control
Who owns the AI workflow after delivery?
Ownership, access and a workable exit, agreed up front. Five questions to ask before anything gets built.
Email
Customer service with a visible route to a person
Escalation as a visible right, not a hidden exception. Four design rules for automated customer service.
Website
Technical SEO in an AI website workflow
The boundary between generation and editorial review, applied to technical SEO. What can be automatic and what cannot.
Website
A website built with AI or by yourself: what do you gain?
What you gain, what you give up and what you must keep doing yourself when AI builds your website.
Documents
AI for accountants: five workflows with control
Five supporting workflows for an accounting firm. Supportive, bounded and each with an explicit review point.
Decisions and control
Reusable AI workflows without client data
Reuse form, not data. What you can copy between clients, and what must stay within each organization.
Decisions and control
Choosing an internal AI pilot: bounded risk
Choose your first AI pilot for learning speed, not the largest promised saving. Four criteria, and why the customer-facing pilot usually waits.
Decisions and control
From AI demo to client process: seven gates
Seven gates between a working demo and a client process that runs every day. Each gate is a question you answer beforehand.
Knowledge, Documents
Connect documents and knowledge without losing permissions
The fastest knowledge base is one that ignores permissions. That is why you build it differently, along four design rules.
Decisions and control
What data does an AI workflow process?
The same eight questions for every workflow. Answer them beforehand, not halfway through the security review.
Meetings, Sales
Meeting information into CRM and tasks
What may flow from a conversation into the CRM is a choice, not a technical detail. Make that choice deliberately.
Decisions and control
Connecting systems: which route should you choose?
The same connection can follow five routes. The right choice depends on permissions, maintenance and how quickly you want results.
Decisions and control
Five AI workflows in professional services
Five familiar workflows, with the review point and the boundary you are better off respecting for each one.
Reporting
Management reporting: AI or BI?
Use AI for the interpretation, not the arithmetic. That keeps your reporting verifiable while still saving work.
Knowledge
Internal knowledge base with AI: sources and permissions
Three things make an internal knowledge base useful: the source, permission and date. If one is missing, usage drops off.
Documents
Document processing: extraction is step one
Extracting the fields almost always works. What follows determines whether document processing actually saves time.
Email
Processing emails with AI: from intake to approval
Not the entire inbox in one go, but four steps you configure, measure and reverse separately.
Meetings
Transcript, summary or action items?
Do not ask for all three. What you need follows from what must happen after the meeting, and each product has a different cost.
Meetings, Project work
From transcript to action after a meeting
A summary nobody opens changes nothing. How a meeting decision actually ends up on a board.
Sales
Updating CRM with AI: fields and rollback
Not every CRM field can tolerate an automatic update. Three categories, each with its control level and way back.
Sales
Sales follow-up without uncontrolled writes
Automating follow-up need not mean write permissions. Keep the CRM reliable while speeding up the work.
Decisions and control
From AI idea to production in five decisions
A pilot rarely fails on technology. These five decision points determine whether an idea reaches production.
Decisions and control
AI, fixed rules or RPA: what fits where?
Choose technology based on the shape of the problem, not the newest term. Three criteria that settle the choice in minutes.
Decisions and control
Which processes can you automate with AI?
Four questions determine whether a process suits AI. In some cases, automation is the wrong answer.