Topic
Buying AI software: diligence, shortlists, and the meter you cannot see
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.
What to look at first
- The billable unit, defined in the contract and not the datasheet
- Audit rights and record retention on consumption data
- Written kill criteria with a date attached
Issues
- Twenty Questions That Separate an AI Vendor From an AI Demo
Every GTM AI vendor demos well. Diligence is how you find out whether the product works on your data, in your process, at your risk tolerance. These are the questions that change the answer.
- What Is AIforce, and How Does It Relate to Agentforce, Claudeforce, and Slackforce?
AIforce is Salesforce's live interface layer that brings CRM data, workflows, logic, and permissions to any AI interface. It sits on top of Data 360, Customer 360, and Agentforce. Claudeforce, Slackforce, and Agentforce Coworker are the first three surfaces. Open beta for Claudeforce is not GA; the product page is stale versus the Sep 15 announcement.
- What Is Salesforce Koa, and Is It GA for Agentforce?
Koa is Salesforce's first CRM reasoning model for Agentforce, post-trained from NVIDIA Nemotron 3 Super inside the Salesforce trust boundary. As of the Sep 15, 2026 press release it is available to select pilot customers only, with general availability expected winter 2026 in U.S. regions. It is a model-provider choice, not a new interface layer and not a Flex Credits usage type.
- Your Buyers Have Already Decided: What to Do When AI Builds the Shortlist Before You Know They Exist
94 percent of B2B buyers use AI during their research and 92 percent arrive with a shortlist already formed. The 61 percent of the journey that happens before a rep is involved belongs to nobody in most orgs. Here is how to take it back.
- Who Audits the Meter on Our AI Agents, and What Audit Rights Do We Need Before We Sign?
Nobody audits it unless you write the right in. On a consumption AI contract the vendor defines the billable unit, detects it with its own system, and holds the only full-fidelity record. Cap vs audit, clause set, parallel count.
- What Kill Criteria Should We Set Before Signing an AI SDR Contract?
Write the fail conditions into the order form before the demo becomes a two-quarter argument. Hybrid 1.9x, reply-rate floors, named redeployment, no headcount-cut ROI story.
Research
Frameworks
Definitions
- Kill Criteria
Kill criteria are the fail conditions written into an AI contract before signature: the metric, the floor, the date it is measured, and the consequence for a miss. Without them a failed pilot becomes a two-quarter argument.
- The Proof Gap
The Proof Gap is money spent on AI with nothing attributable behind it. Tools were bought, pilots ran, time savings were reported upward, and revenue still cannot be tied to any of it. The Revenue AI Report exists to close it.
Open data
- Vendor Cost Per SQL
Cost per sales-qualified lead by AI vendor category, compiled from panel-submitted spend and CRM-verified SQL counts. Methodology, schema, and CSV access. Free download, no signup, CC BY 4.0.
- The Tool Saturation Map
AI vendor density by revenue category: how many vendors compete in each seat and motion, and how the count is moving. Methodology, schema, and CSV access. Free download, no signup, CC BY 4.0.
- The Reversal Ledger
The full Reversal Ledger dataset. Every named AI rollback, shutoff, or reversal by a B2B revenue team, with reason code and disclosed cost. Downloadable CSV, CC BY 4.0. Free download, no signup, CC BY 4.0.
Playbooks
- AI-augmented deal desk (L4)
L4 Orchestrated. Deal desk uses agent to summarize the deal, flag risks, suggest terms. Reduces approval cycle from days to hours.
- Agent observability + cost monitoring (L4)
L4 Orchestrated. Every AI call logged with cost, latency, prompt, output. Without this, you cannot scale agents responsibly.
- RAG-powered competitive intel (L3)
L3 Integrated. Crawl competitor sites, G2, Reddit, earnings calls → embed → reps query in Slack. Beats stale battle cards.