Resources
Ideas and guides with criteria.
Guides and criteria on data, automation and applied AI — to decide better, not to pile up tools.
Foundational reads
Start here.
Three pieces that explain the criteria we work by. If you only read one thing, make it one of these.
- AutomationBefore you automate: why the problem usually isn't the toolsAutomating a messy process scales the mess. How to tell a tool problem from a system problem, and when it's not time to automate yet.5 MIN
- Applied AIWhat applied AI actually means (and doesn't) for a non-technical businessWhat applied AI is and isn't: concrete, measurable use connected to real processes. How to tell it apart from hype, demos and chatbots.5 MIN
- DataScattered data across spreadsheets, your CRM and your ERP: building a single source of truthSpreadsheets, CRM and ERP telling different versions of the business. How to unify your data into a single source of truth: which one rules, who owns it, how to decide.5 MIN
The editorial rule
What gets published here — and what doesn't.
No product news, no tool comparisons. Every piece answers a decision someone has to make — automate or not, unify or not, trust a dashboard or not — and gives the criteria to make it, with its conditions and its limits.
We publish when there is something to say, not on a calendar. An article that repeats what another one already says doesn’t get in — just as a section that doesn’t earn its place doesn’t make it onto a page.
The collection
The collection, by topic.
Foundations
- Process, system, tool: three words that don't mean the same thingProcess, system and tool are not synonyms. Naming the problem wrong leads to buying software for process failures: what to assess before you change anything.5 MIN
- Technical debt for non-technical leaders: the loan nobody wrote downTechnical debt is not a programmers' problem: it is a management decision. What it is, how it builds up quietly and the signals that it is time to pay it down.5 MIN
- What a system integration is (and when it earns its cost)What integrating two systems actually means, why human copy-paste is already an invisible integration, and when connecting them earns its cost.5 MIN
- What is master data (and why the same customer shows up three times)Duplicate customers and one supplier under three names: what master data is, which record wins when two don't match, and who gets to decide that.5 MIN
- Where does this number come from? Minimal traceability for your reportsA number nobody can trace back to its source is a claim, not data. How to build minimal traceability for your reports without making it a technical project.5 MIN
Data
- A shared language for your metrics: making "sales" mean the same thing everywhereSales and finance report different numbers for "sales" and nobody is wrong. When to write down what each metric means and the minimum every definition needs.5 MIN
- When a core system fails: the minimum continuity your operation needsIf your core system won't start tomorrow, what can your company still do? How to decide your minimum continuity: tested backups, hidden dependencies, limits.5 MIN
- Cleaning data: when it's worth it and when it's wasted effortCleaning your whole database chases a state that doesn't exist. How to decide which data gets cleaned, which gets archived and which you let die, guilt-free.5 MIN
- GDPR and your CRM: the operational minimum every SME should have in placeThe operational minimum an SME should have in place for personal data in its CRM: why you hold each record, who can access it and how long you keep it.5 MIN
- Who owns a piece of data? Lightweight governance for SMEsData without an owner degrades as it gets used. How to name an owner for each critical piece of data, what they decide, and how to set it up without committees.6 MIN
Automation
- Automation with people in the loopWhere to place human review in an automated process without slowing it down: impact and ambiguity as the criteria, and which checks are pure theatre.5 MIN
- Knowing when an automation failsA failed automation looks exactly like a working one. The minimum controls — alerts, a log, an owner — so you find out before your customer does.5 MIN
- Lightweight automation: what you can automate without opening a projectWhat you can automate with the tools your company already pays for — email, spreadsheets, your ERP — and how to tell when a small rule stops being enough.5 MIN
- The real cost of an automationBuilding an automation is the cheap part. The full cost structure — dependencies, knowledge, operation, exit — to decide whether it truly pays off.5 MIN
- What to automate first in admin workHow to decide which admin task to automate first: high frequency, clear rules and a low cost of error — and which tasks should stay with people.5 MIN
Applied AI
- An internal assistant on your own documents: what needs to be in place firstAn internal assistant answers only as well as your documents allow. What must be solved — currency, versions, owners — before deciding to build one.4 MIN
- The EU AI Act: a practical guide for SMEsThe EU AI Act is in force and you don't need a legal megaproject to respond: inventory your AI uses, place them by risk level and keep minimum evidence.9 MIN
- Generative AI in admin and customer service: where to startWhere generative AI already pays off in admin and customer service: drafting, summarising, triage and first replies, and the conditions each one needs.5 MIN
- What not to paste into an AI chat (and what to do instead)Your team already uses AI chat tools. Which data must never leave the company, what to do instead, and how to write a short internal rule people follow.4 MIN
- Why AI makes things up (and how to work with it)Why an AI model confidently writes things that are false, where that barely matters, where it is unacceptable, and how to work with the limitation.5 MIN
- Why your team isn't using the AI you boughtYou bought an AI tool and hardly anyone uses it. Before blaming the product or buying another, decide whether the tool failed or the rollout did.4 MIN
Reporting & BI
- Automating your monthly reporting: what to solve firstAutomating a bad report just makes it faster. What must be solved before automating monthly reporting, what to automate and what to keep with people.5 MIN
- Real-time data: when you actually need itReal-time data is expensive and rarely necessary. How to choose an update frequency based on the decisions your data feeds, not on what sounds modern.5 MIN
- Reports nobody reads: pruning your legacy reportingReports produced out of habit: how to decide which ones to retire, an honest way to test it, and what the reports that stay must be able to answer.5 MIN
- The spreadsheet holding your business together: when it stops being a solutionThe heroic spreadsheet is a solution until it becomes a risk. Three signals to decide when to treat it as one, without bans or blind migrations.4 MIN
- How to build a leadership dashboard people actually use to decideHow to build a leadership dashboard people actually use: a handful of well-defined indicators, with an owner and a source, instead of a decorative panel.7 MIN
- A pretty dashboard isn't the same as useful informationA pretty dashboard doesn't guarantee useful information: the usefulness lives in the definitions, the source, the owner and the decision behind it.3 MIN
Decision criteria
- Before you replace your ERP or CRM: signs the tool isn't the problemReplacing your ERP or CRM migrates the mess if the real problem is process or data. Signs the tool isn't to blame, and when a switch is actually justified.5 MIN
- Build in-house or bring in help: deciding who does your data and AI workData and AI: what your current team can solve, what training fixes, and when it pays to bring in outside help. Honest signals in both directions.5 MIN
- Documenting the minimum: so a holiday doesn't stop the businessWhat has to be written down — access, critical processes, decisions — so the business doesn't depend on one person. Minimum viable documentation.6 MIN
- How to evaluate a data or AI proposal before you sign itThe questions to ask before signing a data or AI proposal: what stays in your company, what dependencies it creates, how success is measured, what if it stops.5 MIN
- Sizing your first AI pilot: small, measurable, with the ending written downHow to size a first AI pilot: a small scope, success conditions written before you start, a clear point where you stop, and a plan for what happens if it works.6 MIN
Let's talk
Reading helps. Deciding with someone across the table helps more.
If something here sounded like your case, tell us about it and we'll come back with a first read.
Tell us about your case
A first 30-minute call with direct senior interlocution — no commitment and no sales pitch.