BlockchainGirl

I’m a big fan of the current @united safety video…I like the Rube Goldberg ball rolling machine. Or maybe I have just seen it too many times in the last six months and have convinced myself it’s enjoyable…

Why is data migration hard? Boiled down to its bare bones, the problem is one of field-level and value-level mapping, ETL (i.e., the actual compute operations), and validation. Failure points are possible at each step. Mapping requires domain expertise and/or intelligent matching schemes, as 1:1 maps are often impossible (e.g., the CoA structures in Oracle are significantly different from those in SAP because the application layer itself has different treatments of the concept). We have seen, however, that this problem is solvable (admittedly, we had to build some complex algorithms over the last year or two). ETL can be challenging, but is fairly tractable. Reliability, compute optimization, caching, etc., are long-standing problems with well-developed solutions. Validation - this, in my view, is the most substantial remaining bottleneck. Received knowledge states that only business stakeholders can sign off on any particular data load, since only they understand the business to its bones. Naturally, the number of stakeholders across a global enterprise could number in the hundreds, and thus validation timelines can stretch into months. In total transparency, I think the (uglier) and real reason for requiring business stakeholder sign-off has less to do with presumed domain expertise and much more to do with forcing accountability. You will consume & operate on this data; therefore, you must sign off on it. If issues slip through, you will suffer from them as a consumer, and I do not want to be blamed. And yet, we are increasingly hearing that the consumers of enterprise data are agents, not business stakeholders. Agents will now drive "zero-day closes". Agents will drive invoice approvals. Agents will drive supply chain operations and procurements. Then, should agents be the responsible stakeholders for business validation? Ontology-grounded agents may in fact be more reliable "vetters" than human validators who can only drive sample-based verification given time & bandwidth constraints. Can we square the idea of "autonomous enterprise" with a reliance on full human-in-the-loop validation mechanisms? We are (for now) accepting this cognitive dissonance, but I do not believe it can hold in the long run. In fact (and we will be able to share details soon), some of our customers have now accepted no human-in-the-loop even in business validation activities. These are not small customers in the slightest. My observation: these customers who have truly embraced trust in agents are those who see the most substantial cost & time reductions in transformation activities. In short, trust is now the bottleneck of agentic acceleration, and those who trust the most benefit the most.

An observation…fragmentation of policymaking and lack of international standardization is driving a tremendous amount of labor inefficiency and unnecessary costs in areas that are often unobserved by the broader public. Global technology rollouts are hampered by the need for fit-gap analyses accounting for local regulations and methodologies. A template built in the US will hardly work for every European or Asian tax & trade regime. Meanwhile, building a global template from first principles is not an easy task. Fundamentally, I see this as a failure of international cooperation and standardization. The problem is only getting worse as deglobalization continues. The only solution I see today that does not involve massive labor investments is the “living ontology” approach. We have seen success with this in several clients living the rollout pain points today. A side effect of this global fragmentation is the empowerment of countries with large, (relatively) unified domestic markets such as the US or China. A single technology platform can service huge consumer bases without needing to account for region-to-region discrepancies. Meanwhile, enterprises navigating multi-regime operations in the traditional way face costs reaching tens of billions of dollars per year in aggregate, possibly more when accounting for slowdowns in productivity-boosting technology rollouts.