The best tool for each job stopped being the best answer

Adam Smith's pin factory is usually remembered for the arithmetic. Ten workers, each doing one part of the job, produced vastly more pins than ten workers each making a whole pin alone. What is less often quoted is the condition attached to it, which is that the gains depend entirely on the parts fitting together without friction. Smith was describing workers in one room, handing physical objects to each other. The coordination was free because it was invisible.

For most of my working life, the equivalent logic in an advice firm was unarguable. Buy the best cashflow tool, the best research system, the best back office, the best risk profiler. Each specialist beat any generalist at its own job, so a firm assembling the strongest individual pieces ended up with the strongest whole. I built my own arrangements on exactly that reasoning, and for a long time it was correct.

That assumption has quietly stopped being true, and the mechanism is the one Smith's condition exposes. The gains from specialisation are real and they are roughly linear: each additional specialist tool is somewhat better at its job than a generalist would be. The cost of coordination is not linear. It rises with the number of joins between systems, and the number of joins rises faster than the number of systems. Four tools is six possible connections. Seven tools is twenty-one.

What that looks like in practice is not dramatic. Nobody experiences a coordination cost. They experience a paraplanner typing a date of birth for the third time, a case waiting for a file that lives somewhere else, a review that cannot be evidenced because the record of it sits in a system the compliance report does not read. Each instance is small enough to absorb and none of them appears on an invoice, which is precisely why the total goes unmeasured for years.

I have become suspicious of my own instinct here, because the instinct was formed in conditions that no longer hold. When a firm ran three systems, the join cost was genuinely trivial and buying the best of each was obviously right. The instinct survived into a world where the same firm runs seven, and it now produces the wrong answer while feeling like good judgement.

The firms I see getting more out of a week than their peers are rarely the ones with the most capable individual tools. They are the ones who accepted a slightly weaker tool somewhere in exchange for removing a join, and who measured the week rather than the feature list. That is an uncomfortable trade for anyone who takes professional pride in using the best available equipment, which is most advisers I know, and it is the trade that is now usually right.

There is a wider version of this that I think applies beyond software. Any system assembled from best-in-class components will eventually spend more on integration than it gains from quality, and the point at which that happens arrives earlier than anyone expects because integration costs are paid in attention rather than money. Attention is the thing nobody budgets for and everybody runs out of.

The honest test is not which tool is best. It is how many times a single client's information gets handled between arriving and being useful, and whether anyone in the firm has ever counted.

Specialisation stops paying at the point where the handovers cost more than the expertise.