Anja Warrot

The through-line

Everything Anja Warrot says circles back to one test: does this actually make the work better? "AI wordt pas interessant als het je werk beter maakt. Minder zoeken. Minder manueel werk. Betere inzichten. Meer tijd voor wat echt telt." [1] The vocabulary repeats almost word for word elsewhere, because it is the whole argument: Claude delivers "iets waar we allemaal te weinig van hebben: tijd. Minder zoeken. Minder knippen en plakken. Minder manueel controleren. En sneller van data naar inzichten die je ook écht kan gebruiken." [7] The threat framing gets dismissed each time it appears. "Niet omdat AI je vervangt. Wel omdat het je tijd teruggeeft voor wat écht telt. Dat is geen hype. Dat is slimmer ondernemen." [10] And again: "Niet om je werk over te nemen. Wel om sneller inzicht te krijgen en betere beslissingen te nemen. Daar zit voor mij de echte meerwaarde." [11]

Underneath the AI work sits an older preoccupation with entrepreneurs actually knowing where they stand financially, and with the willingness to question how things are currently done. The two connect: the tooling is a means to the insight, and the insight is a means to a better decision. "De technologie is klaar voor betere inzichten en meer efficiëntie. Ben jij er ook klaar voor?" [8]

On why most people use a fraction of what AI can do

The adoption statistic does not impress her. "81% van de Belgische KMO's gebruikt AI. En toch merk ik dat veel ondernemers er vandaag nog maar een klein stukje van gebruiken. Terwijl er zoveel meer mogelijk is." [11] The specific limitation she names is treating AI as a writing and image toy: "Werk jij vandaag met AI vooral om eens een tekstje te schrijven of een afbeelding te maken? Dan gebruik je nog maar een héél klein stukje van wat vandaag mogelijk is." [7] Her response is to lower the barrier rather than argue, releasing a set of prompts and skills for free with an invitation to "Probeer ze eens uit. Speel ermee. Laat je verbazen." [7]

The other objection she meets head-on is time. "'Ik heb daar eigenlijk geen tijd voor.' Net daarom is dit zo interessant." [10] The shortage of time is the reason to start, not the reason to postpone.

On picking one tool and going deep

Her advice on tool choice is deliberately narrow. "Kies één AI-tool en leer er echt goed mee werken. Dan ben je al veel verder dan de gemiddelde gebruiker. Wij kozen voor Claude." [9] The payoff she promises is concrete and financial rather than generic: "Geen geknutsel meer in Excel" [9], turning figures into "een duidelijk antwoord op de vraag: Ben ik goed bezig?", checking faster whether data are correct and complete, and building insights and dashboards "zonder code of IT-afdeling" [9]. The same list recurs as the target of the model testing: a report that no longer costs half a day, a visual dashboard [15].

On testing every new model against real client work

She treats model releases as a standing obligation rather than news. "𝙇𝙚𝙫𝙚𝙣𝙨𝙡𝙖𝙣𝙜 𝙡𝙚𝙧𝙚𝙣, het staat ook deze week op de agenda." [15] The reaction to an Anthropic launch is honest about the cost: interesting, and "daar gaat mijn planning weer" [15]. Four new Claude models arrived since the start of June, Opus 5 among them, and each triggers the same protocol: give the model an identical assignment again and compare. "Maakt dit model minder fouten? Ziet het sneller verbanden? Bouwt het een beter dashboard?" [15] The decisive question is never technical. "En vooral: 𝙬𝙖𝙩 𝙝𝙚𝙚𝙛𝙩 𝙚𝙚𝙣 𝙤𝙣𝙙𝙚𝙧𝙣𝙚𝙢𝙚𝙧 𝙝𝙞𝙚𝙧 𝙢𝙤𝙧𝙜𝙚𝙣 𝙚́𝙘𝙝𝙩 𝙖𝙖𝙣?" [15] Only what survives that test earns a place in the training and in further optimisation for clients [15].

On what AI can do with an unexplained file

Her sharpest demonstration is a deliberately unhelped experiment: a DWG file of a house handed to Claude with no explanation of how the client worked, no example of his meetstaat, no instructions about his way of calculating, and a request to produce a meetstaat from it [4]. The result impressed her. Part two mattered more. Given the client's own meetstaat for comparison, Claude picked up conventions it had not inferred from the plan alone, and understood that for a contractor and certain goods m³ or kilos were more practical than pure m² [4]. Then it went further and found a human error, because the brick calculation did not match the m² of façade. "Zowel de aannemer als ikzelf waren onder de indruk!" [4] The episode is her evidence that the value sits in verification and pattern-finding, not in generating text.

On the week after the training day

She is blunt that the event itself is the easy part. "De opleidingsdag zelf is zelden het probleem. De week erna begint het echte werk." [2] People leave Reken op Claude with knowledge, concrete applications and appetite, and then face the real question: "hoe zet je AI daarna écht zinvol in binnen jouw bedrijf? Wat levert het je op?" [2] Her answer is a free follow-up evening, the COBIZ AI-Café at the COBIZ-Center in Dendermonde, built around four questions put to the room: what are you using AI for today, what time saving or insight does it give you, where are you still getting stuck, and how can we help each other further [2]. She adds updates on relevant developments and treats it as a networking evening as well [2]. The training itself runs full days from 8.30 to 17.00 with limited places, in Dendermonde and Mechelen, and carries a score of 8,8/10 [9].

On cash, profit and the four C's

The financial argument is stated as a distinction most entrepreneurs blur. "Winst is een cijfer op papier. Cash is wat er effectief op je rekening staat. En daartussen zit soms een wereld van verschil." [13] COBIZ has accumulated its own C's, Consulting, Class, Community, Center, and she is happy to admit it "is een beetje uit de hand gelopen", but the important C belongs to the client and sits on their bank account [13]. Every one of the four is held against a single light: what it yields for your cash, your time and your Comfort [13]. That impatience with paper answers is the same instinct that makes her want figures to answer "Ben ik goed bezig?" [9]

She is also clear about why so many entrepreneurs avoid the subject. Many feel ashamed of struggling with financial questions because they think they should have it under control, when finance is a specialty like any other trade [16].

On questioning how things are done here

She names the phrase she considers most dangerous in a business: "zo doen we dat hier nu eenmaal" [3]. Progress, on her account, rarely starts with something big, but with daring to question one habit, process or way of working, or improving a little every day [3]. That belief has a costly precedent in her own history. Running a data-driven recruitment agency, she and Dirk Colman watched the number of client conversations needed per successful hire double during the war for talent, and the market stayed there, which broke the business model [16]. A diversification plan was already fully worked out, with brand names and logos ready, and they dropped it on holiday after asking themselves: "Kunnen we onze issues waar dat we vandaag mee worstelen — gaat Van A naar Beter eigenlijk niet beter af zijn in een groter geheel?" [16] The no-cure-no-pay model's pressure and market consolidation drove the sale of Van A naar Beter to Select Group after ten years, and the reasoning included the reinvestment treadmill: "Elke euro die binnenkomt die ga je terug investeren in de nieuwe mensen." [16]

On handling a handover honestly

The mechanics of announcing the sale are where she draws a transferable lesson. Select's people were on the parking lot the same morning the sale was announced, so there was no gap between the news and the buyer's presence, and that plus honest answers to questions let the team switch mentally within a day [16].

On clients, trust and the coffee invitation

Her account of the work is relational as much as technical. Being trusted by clients to work on crucial steps in their business, and to take steps together that make the company genuinely better and stronger, is a compliment she thinks too little about day to day, and she credits Elien Defraeije with pointing it out [6]. "We staan er zelf niet vaak genoeg bij stil hoe bijzonder dat is. Vandaag dus wél." [6] The entry point she offers is correspondingly low-key: a look at the training offer and an open invitation, "Zin om samen een koffie te gaan drinken? We bekijken het samen." [12] The same door is left open after the free prompts: if you then wonder what this could mean for your own company, "Dan weet je mij te vinden." [7] The peer instinct extends beyond clients, to recommending a company and its open job [14], and to a week in Swedish Lapland with Ben and a group of entrepreneurial people, "Heel benieuwd wat een week Into the Wild gaat brengen." [5]

Takeaways

• Judge any AI application by one standard: less searching, less manual work, better insight, more time for what matters [1][7].
• Do not spread yourself across tools. "Kies één AI-tool en leer er echt goed mee werken" puts you ahead of the average user [9].
• Using AI only for text and images means using a very small fraction of what is possible today; the payoff sits in data, verification and dashboards without code or an IT department [7][9][11].
• Handed only a DWG file with no instructions, Claude produced a meetstaat, learned unwritten conventions from a comparison, switched to m³ and kilos where more practical, and caught a human error in the client's own brick calculation [4].
• Test every new model by re-running an identical assignment and asking what an entrepreneur actually gains from it tomorrow [15].
• The training day is rarely the problem; the week after is, which is why follow-up sessions like the COBIZ AI-Café exist to share applications, time savings and sticking points [2].
• Profit is a figure on paper, cash is what is actually in the account, and the gap between them can be a world [13].
• When a market shift doubles the effort per result and breaks the model, a worked-out diversification plan is worth abandoning if the company, clients and staff are better off inside a larger group [16].
• On announcing a sale, leave no gap between the news and the buyer being physically present, and answer questions honestly; a team can switch mentally within a day [16].
• The most dangerous sentence in a business is "zo doen we dat hier nu eenmaal" [3].