Some notes on where data offices are headed

A few things I've been kicking around. Not fully baked, but worth getting down before I lose the thread.

Logical positivism is the language of our field

Logical positivism, or logical empiricism if you want the fuller name, has been on the outs in philosophy since Popper punctured it. But in a world where stakeholders just want to know what's real, verificationism is still the working language of the job.

We want to point to a system of record every time we put out an analysis. That instinct is out of step with how the rest of modern discourse operates now, which runs almost entirely on falsificationism: don't tell me it's true, tell me how you'd know if it were false. Data offices are one of the last places still speaking the older dialect, and I don't think that's an accident. Verification is what the job actually requires, whatever the philosophers moved on to.

The mandate has become too unwieldy

The mandate for data offices is too large as it stands. My thesis: data offices took on governance as the bargaining chip that earned us a seat as our own office. The C-suite never wanted to own governance directly; the mandate is too big, and the world economy is too far behind on it to make it anyone's clean win. Analytics and data science outpaced governance years ago, and now governance eats an outsized share of our time just trying to catch up.

Data offices accepted governance to round out a complete mandate but only have real competency in maybe a third of the ground it covers. IT and Legal hold most of the rest. Layer AI on top of that and you've handed data offices a transformation mandate too, on top of everything else. Staff now have to maintain what already exists and build what's new, at the same time, with the same headcount.

Consider life on the ground in applied data management: let’s say I have five staff members who work 40 hours per week; meaning, I have 200 total hours to allocate per week, independent of mandate. My job as a leader is to identify problems and trends, strategize, prioritize and, where necessary, allocate work to time. Two hundred hours becomes my workspace unless I am willing to cede operational expertise and just pay for ongoing contractual bailouts, which is a horrendous practice and leaves your office threadbare for skill development.

Data analytics is heading for a crack-up in the next 3 to 7 years

Data offices absorbed AI, but they've had to fill the gaps with outside applications and outsourced development. Some shops have built real internal competency. Most haven't.

When quantum hits the general economy, it's going to make the AI boom look quaint by comparison. AI and ML still run on traditional von Neumann architecture. Quantum doesn't. It requires an actual change of perspective, not just new tooling, and a substantial re-education for anyone who wants to work in it. Quantum speaks in logic gates, entanglement, and wave function collapse, and none of that factors into how we talk about data today.

Data offices are going to lean on outsourcing even harder than they do now. The real risk sits with in-house data scientists, who are at real risk of becoming glorified contract managers if they don't get ahead of the shift.