AI & Technology
The economic and business consequences of the technologies changing how work gets done.
What's happening
Capable systems are being embedded into ordinary business workflows faster than most organisations have built the processes to supervise them. The technical frontier and the operational frontier have come apart.
Why it matters
The interesting question is not what a model can do in a demonstration but what it changes in a cost structure, a job description and a liability. Value shows up where a task is genuinely repetitive, verifiable and expensive — and evaporates where verification costs as much as the work.
Forces behind it
- Unit economics
- Inference costs money per use. Products whose margin depends on that cost falling are making a forecast, not a plan.
- Capital spending
- Compute, power and data-centre capacity are long-lived, heavy commitments made against demand that is still being discovered.
- Verification cost
- The expense of checking an output. Where it approaches the cost of producing the work, the automation saves little.
- Enterprise adoption
- Procurement, security review, integration and retraining move on institutional time, which is why capability and deployment diverge for years.
- Labour composition
- Tasks are affected before jobs are. Roles are reassembled around what still requires judgement, context and accountability.
- Regulation and liability
- Rules on data, disclosure and responsibility for automated decisions determine which use cases are commercially available at all.
What to watch next
- Cost per unit of useful output, not cost per token
- Whether adoption is expanding beyond pilots into core workflows
- Capital expenditure commitments relative to disclosed demand
- Where verification is cheap, and where it quietly is not
- Power and infrastructure as the binding physical constraint
- Tasks being reassigned versus headcount being changed
- Regulatory posture on disclosure, data and accountability
AI & Technology articles
The Part of AI That Has a Bill Attached
Every response a model produces consumes compute that somebody pays for. That single fact constrains which AI products can work as businesses.
Automation Is Cheap Until You Have to Check the Work
The decisive variable in most automation decisions is not whether a system can do the task. It is what it costs to verify that it did.
How AI Is Actually Changing the Way Businesses Operate
Past the hype cycle, a quieter shift is underway in how companies use data, make decisions, and structure work.
Tools worth knowing about, when we find them.
We are building a curated shelf of financial tools and resources we genuinely believe are useful. Nothing fake, nothing paid-for-placement dressed up as a recommendation.
Partners — Coming Soon