RiseLab

Go/No-Go Decision: How to Gate Pilots and Bids on Evidence

Most stalled AI pilots and weak bids share one root cause: nobody forced a real go/no-go decision. Here is a gate built on pass/fail criteria and evidence-backed scores, and where RiseLab produces that evidence.

RiseLab ·

A hand placing a signed decision card on one of two paths, labeled proceed and stop, beside a checklist of pass/fail gates with evidence documents attached

A go/no-go decision is the moment someone with authority commits to proceed or stop, based on evidence measured against criteria agreed in advance. Most teams skip it. The AI pilot drifts past its review date because no one owns the stop call. The RFP gets a bid because the deadline is close and nobody wants to be the person who said no.

This article lays out one gate that works for both situations, and shows where RiseLab produces the evidence the gate depends on.

Why the gate gets skipped

Three things usually go wrong, and they have little to do with the scoring math.

  • No named owner. A committee gathers opinions, but no one signs the verdict, so the decision never closes.
  • Evidence is scattered. Security notes sit in one folder, past performance in another, test results in a chat thread. Nobody can assemble the case in an afternoon.
  • Enthusiasm stands in for proof. The team likes the pilot or the opportunity, and the scoring session turns into a debate about feelings.

Fix the owner problem with a single accountable signature. Fix the other two by deciding up front what counts as evidence.

Split pass/fail gates from weighted scores

The most useful design choice is separating criteria you can trade off from criteria you cannot.

Weighted scores suit things like expected value, resourcing, and strategic fit. A strong result on one can offset a weak result on another.

Pass/fail gates suit everything else. If a gate fails, the answer is no-go and the weighted total is irrelevant. Set the weights and gates before anyone scores, and write down what each score level means.

For an AI pilot in a regulated setting, the gates are usually questions like these, and each should be answerable with an artifact rather than a promise:

  • Does sensitive data get stripped before it reaches a model? RiseLab redacts sensitive data at ingestion, before it reaches a model.
  • Can you show who accessed what? The RiseLab Platform supports SSO, role-based access control, audit logs, and encryption at rest and in transit.
  • Do consequential outputs pass a person first? It routes critical AI decisions through configurable human approval workflows, with an audit trail for every approval.
  • Has the system been attacked on purpose? It tests models against adversarial attacks, including jailbreak and prompt injection testing and bias and toxicity probes.
  • Can a reviewer inspect the vendor's own results? RiseLab publishes a public trust scorecard that runs a fixed 94-probe adversarial corpus across 8 categories against its deployed input filter, and the same corpus gates every release.

You can review all of this on the RiseLab Platform page.

Score your own evidence against a framework

Gates tell you what must be true. You still need to know how close you are. RiseLab's compliance tooling scores an organization's own evidence against framework templates, including NIST CSF 2.0, ISO 27001:2022, the HIPAA Security Rule, FAIR, and contract and audit review, and ranks coverage gaps by impact. A score is not a certification, so treat it as a way to see which gap to close first, not as a pass stamp.

For ongoing obligations, RiseLab tracks HIPAA, GDPR and SOC2 compliance with automated reports and regulatory monitoring. It also provides a policy template library with audit trails. That means the evidence behind a gate decision can come from a system you already run rather than from a scramble the week before the meeting.

Make every score trace to evidence

The rule that keeps scoring honest is simple: no document, test result, or named sign-off, no high score. A score justified by "we feel good about it" should never rise above the middle of the scale.

RFP Copilot applies this rule to proposal work. It builds an evidence graph from a company's own material, and nothing becomes corporate fact until a person confirms it or a public record establishes it. In October 2026 it began pulling a company's public USAspending awards and SAM.gov registration into that graph each night, with no upload and no AI generation involved.

When evidence is missing, it does not wait for someone to notice. RFP Copilot raises evidence requests automatically for gaps such as expiring credentials or high-severity gaps on active pursuits, each assigned to an owner and due by the relevant deadline. The same discipline applies to drafts: a grounding audit flags any claim without supporting evidence.

The bid/no-bid version of the same gate

A bid decision uses the same structure with different rows. The pass/fail gates are eligibility, mandatory requirements, and whether you can deliver what the solicitation asks. The weighted criteria are fit, competitive position, and the cost of pursuit.

Here is how RFP Copilot supplies evidence for each row:

  1. Find and read the opportunity. It discovers opportunities from SAM.gov, Grants.gov, Govly, TX SmartBuy, World Bank and USAspending re-competes in one intake. It shreds an uploaded PDF or DOCX solicitation into a structured requirement set with mandatory flags, evaluation criteria, submission format and attachments.
  2. Score fit and record the call. It scores opportunity fit against a company profile and records a Go/No-Go decision.
  3. Check feasibility. Since October 2026 it checks staffing and schedule asks against what is on record, rating each red, amber or green with what would close any gap.
  4. Surface risks before you bid. The pre-submission risk register detects contract clauses and known risks, and some carry a walk-away trigger.
  5. Test your price. A price corridor compares your estimate with past federal awards and flags a possible winner's curse.
  6. Read the buyer. A buyer evaluation dossier puts GAO bid-protest decisions on record against the buyer.

For the overall readiness read, RFP Copilot shows win readiness as a band with drivers tagged KNOWN, INFERRED or UNKNOWN instead of a win-probability percentage. That framing helps a gate meeting. A pile of UNKNOWN drivers is a task list, not a reason to guess. It also separates capability gaps from evidence gaps and suggests teaming candidates from public award data.

Capacity is part of the decision too. Win Plan now builds a roster of named people with monthly hours, and every pursue decision lists the lower-effort items it displaces. A bid you cannot staff is a no-go, however attractive the contract looks. Details are on the RFP Copilot page.

Use a conditional go with an expiry

Most stalled work deserves neither a clean go nor a clean no-go. It deserves a conditional go: name the blocking condition, the owner, and the date. If the date passes without the condition met, the decision converts to no-go without another meeting.

Some of that discipline can live in the tooling. RFP Copilot's Pursuit Agent reads gaps and open asks on each live pursuit and proposes proceed, team, shelve or wait with reasons tied to sources, and every proposal is decided by a person. Pricing and contacting the contracting officer are never agent actions. Pursuits past their deadline are proposed for shelving regardless of readiness. The agent never takes the decision away from the accountable owner; it keeps the clock visible.

Close the loop after the decision

A gate is only as good as what you learn from it. Two habits make the next one sharper.

First, record what you expected. RFP Copilot logs predicted concerns at submission and grades them against the real debrief in a prediction ledger. It also schedules post-submission follow-ups, including the FAR 15.505/15.506 debrief windows.

Second, track what you promised. The Commitment Ledger captures the commitments made in a submitted proposal as individual items, and a person marks each as met, at risk, missed or dropped. Attribution analysis is withheld until an organization has 25 wins and 25 losses, and the product states that gate openly rather than showing an effectiveness score. That restraint is the right instinct for your own gate: do not claim a pattern the data cannot yet support.

For AI pilots, keep the approval trail. Every approval in a RiseLab human-review workflow carries an audit trail, so the record of who decided what, and on what evidence, is already there when the next reviewer asks.

Next step

Pick one stalled pilot or one live bid and write down its pass/fail gates today, before anyone scores anything. Then check how many of those gates you can answer with an artifact instead of an assurance. If you want to see how the evidence side works for AI pilots, start with the RiseLab Platform. If your gate is a bid decision, walk through a live pursuit in RFP Copilot.