Reference
Topics
Each topic collects the issues, research, frameworks, and definitions behind one decision, with the rule we apply when the evidence is read together.
AI SDR
Should we buy an AI SDR, and how do we know it worked?
The seat price is the smallest line in the model. The published work on this site covers the full cost of an AI SDR, the fail conditions to write into the contract before signing, and the truth tests that separate real pipeline from activity.
Decision rule. Do not sign until the fail conditions, the redeployment plan, and the pipeline definitions are written down and owned by a named person.
4 issues · 5 research themes · 2 frameworks · 2 datasets · 4 playbooks
Read the evidence →AI pricing and contracts
How do we buy and forecast AI spend we cannot predict?
Pricing has moved off the seat, and the meter now sits between the vendor and your forecast. The issues below cover seat, consumption, and outcome pricing, credit models that bill in arrears, and deals that require two contracts and two meters.
Decision rule. Never sign a meter you cannot forecast at month-end. Model the invoice before the pilot, and name the person who owns each meter.
15 issues · 3 research themes · 1 frameworks · 2 datasets · 2 playbooks
Read the evidence →CRM data readiness
Is our CRM ready for AI agents?
AI on a dirty CRM produces a more confident version of the same error. The material here covers the readiness assessment to run first, the definitions and freshness checks that come before a pilot, and the evidence on what conditions agents need to hold up.
Decision rule. Fix definitions and freshness before you fix tooling. If two dashboards disagree today, an agent will disagree faster.
4 issues · 3 research themes · 2 frameworks · 2 datasets · 4 playbooks
Read the evidence →AI ownership and governance
Who should own AI across the revenue organization?
AI in go-to-market usually sits between RevOps, IT, enablement, and whichever leader moved first, which is why wins stay inside the room that built them. The work here sets out an ownership model, the failure pattern when nobody owns it, and a read on each revenue seat.
Decision rule. One named owner per system, per meter, and per decision. Shared ownership is the same as no ownership once the pilot ends.
4 issues · 4 research themes · 2 frameworks · 2 datasets · 3 playbooks
Read the evidence →Measuring AI ROI
How do we prove what AI actually returned?
Most reporting measures effort. The material here covers the measured distance between AI spend and attributed value, where recovered hours disappear, the OAR Matrix for diagnosing which side of the value line you are on, and how to report the result.
Decision rule. Report the decision the number changes. If a metric cannot change a staffing, spend, or process decision, it does not belong on the board slide.
4 issues · 3 research themes · 2 frameworks · 3 datasets · 2 playbooks
Read the evidence →Enablement and adoption
How does enablement change once agents do the execution?
Execution stops being the constraint and judgment becomes it. The work here covers what changes for enablement, how just-in-time coaching becomes measurable, a review method for AI output, and the research on how adoption actually moves.
Decision rule. Train the review, not the tool. If a rep cannot say why an output is wrong, the deployment is not adopted.
4 issues · 4 research themes · 1 frameworks · 2 datasets · 4 playbooks
Read the evidence →AI content quality
Is AI-written content still working, and how do we keep quality up?
Output volume rose across the category at the same time, so the marginal piece bought less attention than it did a year earlier. The material here covers the saturation evidence, an editing method that catches the failure modes, and the argument for pointing models at actions rather than more copy.
Decision rule. Ship the piece only if a named person will defend it. If nobody will attach their name, the model wrote it for nobody.
4 issues · 3 research themes · 1 frameworks · 1 datasets · 3 playbooks
Read the evidence →Headcount, quota, and comp
How do quota, comp, and headcount change once AI carries part of the work?
Several public headcount cuts tied to AI were reversed inside a year, which is the cleanest available evidence that capacity claims outran delivery. The material here covers those reversals, the argument for treating AI as lift rather than reduction, and what changes in quota and comp design when it is real.
Decision rule. Change the comp plan only after two quarters of held capacity. Reversals cost more than the payroll they saved.
4 issues · 5 research themes · 2 frameworks · 2 datasets · 3 playbooks
Read the evidence →Customer success and renewals
Where does AI actually help customer success and renewals?
Post-sale teams sit on structured usage and support data, which is exactly what models need, so the signal is cleaner here than in outbound. The material here covers renewal risk detection, what the seat-by-seat read shows for CS, and the working patterns for health scoring, onboarding, and save motions.
Decision rule. Score risk only if a save motion is already staffed. A health score with no owner is a dashboard, not a program.
4 issues · 3 research themes · 2 frameworks · 2 datasets · 4 playbooks
Read the evidence →Pilot to production
Why do our AI pilots never reach production?
Pilots are graded on whether the output looks right, and production is graded on whether a process changed, which are different tests. The material here covers the sequencing decisions that get skipped, where recovered hours disappear, and what the adoption evidence shows about which deployments survive.
Decision rule. Name the process the pilot replaces before it starts. A pilot with no retired step becomes an extra step.
4 issues · 3 research themes · 2 frameworks · 2 datasets · 3 playbooks
Read the evidence →Buying AI software
What do we ask before we sign an AI vendor?
On a consumption contract the vendor usually defines the billable unit, measures it with its own system, and holds the only full record, so the questions that matter are about counting and stopping rather than features. The material here covers the diligence set, how shortlists get built, and the audit rights to write in before signature.
Decision rule. Ask who can stop the spend at two in the morning. If the answer is the vendor, you have not finished negotiating.
6 issues · 3 research themes · 2 frameworks · 3 datasets · 3 playbooks
Read the evidence →GTM architecture
How should we structure the GTM stack and org for agents?
Agents call systems directly, which makes the interface layer optional and the data layer decisive, and it puts pressure on org charts built around handoffs. The material here covers headless stack design, moving from hierarchy to graph, loop engineering, and what each revenue seat looks like under that model.
Decision rule. Design the loop before the tool. If work still crosses four handoffs, an agent just makes the handoffs faster.
6 issues · 4 research themes · 2 frameworks · 2 datasets · 4 playbooks
Read the evidence →Shadow AI
What do we do about AI tools our team adopted without us?
The most common AI result in a revenue org is a private productivity gain that never becomes a team capability, which looks like adoption and returns nothing at the org level. The material here covers that pattern, what the sprawl and trust evidence shows, and the sanctioned alternatives that pull usage back into view.
Decision rule. Sanction faster than you ban. Every week a good tool stays unapproved is a week the work moves off your systems.
4 issues · 3 research themes · 2 frameworks · 2 datasets · 4 playbooks
Read the evidence →