Research library
Bundled reports.
Our Insights, grouped into four themed reports you can read end to end or hand to a committee. Unlock any report with your work email.
The Eval Discipline Field Guide
Measuring whether enterprise AI actually produced anything
A consolidated report on why most enterprise AI programs never produce a measured outcome, what a defensible eval looks like in production, and how to choose metrics that survive contact with real operations.
- Why adoption metrics quietly replace outcome metrics — and how to stop it
- Baselines that can actually beat you, and how to set them before you build
- The failure condition test: naming the losing number before the project starts
AI Governance & Agent Risk
Where the control boundary belongs when software acts on your behalf
A practitioner's report on delegation boundaries, prompt injection as a capability problem, connector risk, browser agents holding live sessions, and how EU AI Act timing lands on operations.
- Treating shadow AI as a requirements document instead of a violation
- Why every connector you approve is a delegation decision
- What changes the moment an agent can spend money
Insurance AI Operations
Claims, underwriting, and the judgment nobody priced
How AI agents change claims decisions, underwriting judgment, and policy servicing — including what breaks when you automate half a job and why brittleness was sometimes doing useful work.
- Why a claim is a decision, not a task — and what that means for automation design
- The hidden cost of automating the easy half of a role
- Where accountability has to sit when an agent participates in a decision
AI Strategy & Economics
Model choice, inference pricing, and how far to delegate
The economic decisions behind enterprise AI: framing the strategy question correctly, choosing the cheapest model that is right enough, and the pricing assumptions most programs underwrote without noticing.
- The strategy question that produces a real answer instead of a roadmap
- Autonomy framed as a delegation problem with explicit boundaries
- Choosing the cheapest model that is right enough for the decision at hand
CX Operations in the Agent Era
Contact centers, complex cases, and the decisions worth automating
A working report on why deflection and handle time mislead, how routing quality determines everything downstream, what full-population quality review changes, and where escalation should stay human.
- Why deflection and resolution diverge exactly where the revenue is
- Replacing AHT targets with cost per resolved issue including downstream contact
- Scoping voice AI as a routing capability before a resolution capability
Construction & Field Operations
Submittal queues, change order evidence, and early schedule-risk detection
A practitioner report on submittal queues as schedule risk, contemporaneous evidence for change orders, and early detection of milestone slips on commercial projects.
- Reading the submittal and RFI log as a decision queue with named owners
- Assembling change order evidence contemporaneously instead of forensically
- Surfacing milestone risk weeks early from information the project already holds
Foundation Repair Operations
Speed to lead, scheduling, and the proposal window
A practitioner report on the three windows that decide a residential structural repair business — the minute after the inquiry, the hour the board is built, and the ninety minutes the inspector is on site — plus the roll-up integration failure that follows.
- Why lead response latency, not sales skill, decides most booked inspections
- Counting drive time as capacity before approving the next inspector hire
- Designing a reschedule cascade that a coordinator can approve in minutes