Be the Algorithm: Run Your Platform by Hand Before You Automate It

There's a stage in every platform's life when the network you're selling doesn't fully exist yet. The directory is thin. The data set is small. The "matching engine" is a founder reading inquiry emails late at night. Every platform passes through this stage. The only real question is whether your clients ever find out.

Andrew Chen gave the survival tactic a name: Flintstoning. Fred's car looked like a car and moved like a car, but it ran on his legs. An early platform should work the same way — complete on the outside, powered invisibly by the founder's own labour underneath.

Run the math on a typical methodology business. With 25 practitioners in your ecosystem, you can deliver excellent work in a handful of specializations and geographies. What you cannot do is match every client inquiry within 48 hours. Certain industries are barely covered. Certain regions are blank. Certain specializations come down to one person who's already committed for the quarter.

Now imagine letting a client see that directly: a directory full of obvious holes, or a matching tool that comes back with nothing. Chen has a phrase for this experience — the "moment opposite of magic." The client concludes the platform is empty and untrustworthy, walks away, and never returns. One bad first impression, permanent damage.

The fix isn't to build software faster. It's to become the machine yourself, personally, until the real machine is ready.

The Expensive Mistake Comes First

Premature Automation: Engineering Interactions Nobody Has Validated

Before the playbook, name the trap — because most founders fall into it before they ever hear the alternative. The trap is premature automation: writing code for interactions that have never been proven by a human doing them manually. More methodology platforms die this way than die from competitors.

The pattern is depressingly consistent. A founder spends EUR 100,000 on a matching algorithm without ever having matched 50 clients by hand. They ship an automated benchmarking dashboard sitting on 30 data points. They launch a self-service practitioner directory with 15 profiles — 8 of them belonging to practitioners who are no longer active.

Then one of two things happens. Either the technology runs flawlessly and nobody touches it, or — worse — clients do touch it and get burned, because no software can compensate for a network that's too sparse to deliver the promise. The algorithm picks the "best fit" from a pool of three practitioners, and none of the three actually fits. The dashboard renders benchmarks built on 12 data points, which mean nothing statistically. The directory serves up 5 profiles that haven't been touched in six months.

Alex Moazed compresses the lesson into five words in Modern Monopolies: "Chase two rabbits, both escape." In your first 12 months there is exactly one rabbit worth chasing — proving the Core Transaction works through human-to-human delivery of your methodology. The technology rabbit waits. You earn the right to chase it only after you've performed the transaction by hand enough times to know precisely what the software should do.

That's why Flintstoning isn't a fallback for founders who can't afford engineers. It's the strategy. Months of manual operation teach you which steps must be fast, which must stay personal, which can be standardized, and which will need human judgment forever. No product spec can tell you that. Only repetition can.

Your Real Job Description During the Cold Start

Six Roles You Play Personally — Each With an Automated Successor

Flintstoning is not deception. The client gets a genuine, high-quality experience; the only thing hidden is how much human effort produced it. Every role below has a future automated version. The manual version exists for two reasons: to prove demand for the interaction, and to protect the experience while the network is still filling in.

The concierge. A mature platform offers self-service assessments, auto-generated reports, and a searchable practitioner directory. Yours offers none of that yet — so you personally guide each client through the assessment, produce their report yourself, and make the practitioner introduction with context prepared for both sides. Every step is manual. Every step feels seamless to the client.

The matchmaker. When a completed assessment lands, no algorithm fires. You read the results, identify what the client is missing, weigh each practitioner's specialization against their current availability, and send the introduction email yourself. The client experiences a tailored match. What actually happened is that you were the algorithm.

The analyst. You don't have 500 assessments per industry segment. You have 50 in total. So you open the spreadsheet and assemble each client's comparison by hand: "Based on the 12 financial services assessments in our database, your score of 2.3 places you in the bottom quartile." Real data, honestly framed, manually compiled. The automated report generator comes later.

The switchboard. Practitioners on a young network don't yet refer work to each other on their own. You make it happen: you hear that James's manufacturing client needs data architecture help, you know that's Sarah's specialty, and you broker the connection. You are manually producing the network behavior that should eventually run without you.

The publisher. One day the platform will generate benchmark reports on its own. Today, you write the quarterly insight piece yourself from whatever data exists: "Across our first 75 assessments, organizations scoring below 2.0 on automation maturity are disproportionately concentrated in three industries." A genuine finding from genuine data — typed by you.

The quality inspector. Eventually, satisfaction surveys trigger automatically after every engagement. For now, you pick up the phone two weeks after each engagement closes: How was it? What should we improve? Would you recommend us to a peer? Tedious, yes — and irreplaceable, because those calls police quality and feed the referral loop at the same time.

Six roles. All performed by hand. All deliberately temporary. Together they sustain the trust of clients and practitioners through the exact period when the network can't yet sustain it alone.

Graduation Day

Three Bottlenecks That Write Your Automation Roadmap for You

Flintstoning has an expiry date built in. As the network grows, the manual load becomes progressively impossible — and each point where it breaks is not a crisis. It's your automation roadmap announcing itself, ranked by urgency, validated by lived experience instead of guesswork.

Bottleneck one: matches slip past 48 hours. Inquiries are stacking up because one person can no longer review and match them fast enough. That's the cue to build the matching algorithm — and you'll build it well, because months of hand-matching taught you exactly which factors carry weight.

Bottleneck two: benchmark demand outruns the spreadsheet. Clients are waiting days for comparisons you assemble by hand. That's the cue to build automated reporting — and you already know precisely which comparisons clients care about, because you've been curating them one at a time.

Bottleneck three: follow-up calls start getting skipped. Engagements are closing without a quality check because there aren't enough hours to call everyone. That's the cue to build the automated survey — and the questions write themselves, because you've asked them personally dozens of times.

Notice what each bottleneck hands you: the what and the how of the next build, specified by repetition rather than imagination. That specification is the asset no product manager could have drafted on day one.

Fred Flintstone's feet weren't the product — they were the proof. Run the platform on muscle until the network is dense enough to deserve a motor. Then build the motor to do exactly, and only, what your own effort already proved clients will pay for.