RFP template & evaluation kit
Everything a network operations team needs to issue an RFP for agentic AI and score the answers on evidence: a Word template you can put your name on, an Excel requirements matrix with weights and must-have gates, a 45-day proof-of-value test plan and the questions to ask references. Vendor-neutral by design · including toward REAP.
What is in the kit
Fourteen sections: instructions to bidders, your background and objectives, current environment questionnaire, requirements by section, evaluation method, proof-of-value stage, commercial and legal requirements, submission format. Placeholders for your company, dates and contacts.
120 requirements in 15 sections. Each carries a priority (must / should / could), a default weight you can change, a response code the vendor must choose (fully supported today · partial or configurable · roadmap with date · not supported) and an evidence column. Scores compute automatically. A must-have answered “not supported” removes a vendor; a must-have answered “roadmap” needs a committed date in the order form and is re-tested in the proof of value, and two or more remove the vendor.
Twenty-two tests mapped to the requirements, each with a pass condition you can observe in your own environment in 45 days: incident replay, change verification, findings review, runbooks under approval, security review.
Twelve questions to ask a vendor’s references, a demo script that keeps the demo on your data, and the scoring rules: how to weight, how to treat “roadmap”, how to break ties.
How to use it
Every requirement describes a practice or a test, not a product. Several can be met by more than one vendor or by a well-run internal team, and REAP’s own answers are scored with the same codes: some requirements we meet today, some in pilots, and where a capability is roadmap we say so, on per-action autonomy modes and outcomes written to the change record, REAP’s answer today is “roadmap”. REAP has design partners and no production references yet; section 1 asks every vendor to say the same plainly. The kit is free to copy and adapt with attribution.
Get the document
Tell us where to send it. You receive one email with a personal link to the kit: the Word template, the Excel requirements matrix, the test plan and the reference questions.
The fifteen sections
| # | Section | Examples of requirements |
|---|---|---|
| 1 | Company, viability and references | Founding team, funding, pilots and design partners, security program, references willing to speak (or why there are none yet) |
| 2 | Coverage and discovery | Vendors and OS versions supported; discovery in days; topology accuracy stated and measured |
| 3 | Organizational context | Runbooks, design documents, tickets and change records ingested and cited in investigations |
| 4 | Incident reasoning and root cause | Evidence-linked root cause on replayed incidents; blast radius; P50/P95 time to answer |
| 5 | Alerting and noise | Prioritization by topology and business context; redundancy-compromised alerts |
| 6 | Change assurance | Pre-execution impact analysis; pre/post verification for changes made anywhere; one outcome per verified change (Verified, Degraded, Rolled back, Inconclusive or Blocked) recorded in the change record |
| 7 | Resilience, posture and compliance | Standby health; backup-path capacity; drift; contextual CVE (opens in a new tab) exposure; evidence packages |
| 8 | Automation, runbooks and autonomy | Natural-language runbooks; approval gates; rollback; autonomy modes per action and role |
| 9 | Transparency and evidence | Every conclusion traceable; uncertainty stated; hypotheses recorded; engineer scoring |
| 10 | Integrations | ITSM read/write, PAM brokering, monitoring, on-call, collaboration, knowledge, identity, controllers |
| 11 | Architecture and deployment | SaaS, customer cloud, on-prem and air-gapped; connector model; scalability; multi-tenancy |
| 12 | Security, privacy and AI governance | Least privilege; data classes; retention; model endpoints and training policy; audit export; disclosure |
| 13 | Measurement and proof of value | Baseline, eight measures, business criterion, decision date, artifacts delivered |
| 14 | Commercials | Pricing model that does not penalize adoption; pilot credit; term; expansion; SLAs |
| 15 | Implementation, support and services | Time to value; enablement; support model; roadmap transparency; exit and data deletion |
Thirty minutes with a solutions engineer: your doors, your weights, your timeline. No pitch.