Verified
Official vendor pricing, plan limits, named integration paths, enterprise controls and policies. Every verified integration should have an exact product/plan scope and source.
Methodology
Autonomy should rise only when evidence rises.
PriceBolt separates workflow understanding, simulated execution, tool use, actual-product execution and controlled live evidence so an SMB never mistakes a good model answer for proof that a business process is safe to automate.
Measured
Independent benchmark results with a named methodology, measured variant and timestamp. Frontier-model scores do not automatically become product rankings.
Observed
Actual prices, usage patterns and outcomes contributed or measured from businesses. Peer medians are withheld until sample sizes are meaningful.
Estimated
Finder recommendations and projected fit. These are modelled judgments, clearly separated from measured ROI.
Workflow mapped
PriceBolt understands the process, context, risk and current human baseline. No AI execution claim yet.
Test defined
Rules, exceptions, hard controls and human escalation boundaries are explicit enough to score.
AI simulated
An AI candidate runs independently audited normal, boundary, missing-information, hard-control and adversarial synthetic cases. Decision-only runs do not claim that drafted replies, reports, extracted fields or other generated artifacts were verified unless those artifacts were actually generated and graded.
Tools tested
AI executes the controlled workflow through tools rather than merely describing what it would do.
Your AI tested
The intended commercial AI setup executes the controlled workflow. Product/plan identity remains explicit and bounded by what PriceBolt actually observes.
Live validated
Controlled real-business cases are processed by the intended commercial AI while PriceBolt directly records the recommendation and human review. This establishes real-work decision evidence, not automatic production-write permission.
Claim rule: evidence never jumps levels—and recommended autonomy is capped by the evidence level. L2 simulation cannot justify autonomous live writes. A commercial-host synthetic run cannot become proof about the SMB's real systems. L5 shadow/approval evidence still keeps production writes behind human approval until the exact system/action path has a PriceBolt-verified write, post-state, least-privilege and rollback check. Critical failures or high-consequence workflows force autonomy back down.
After L5 — production write readiness: this is a separate scope-specific control, not a new marketing evidence level. Manual or customer-reported connector evidence can document implementation context but cannot clear the gate. Only a configured PriceBolt-verified adapter can mark the named system/action scope ready after observed write and rollback evidence.
01
Integration evidence
Unknown is not unsupported.
PriceBolt records how a named system connects: native, official/directory connector, hosted/custom MCP, synced connector, API, or another verified path. Generic MCP support alone does not prove compatibility with every SaaS, and absence of a named record is never converted into a negative claim.
02
No pay-to-rank
Vendors can buy visibility later. They cannot buy the answer.
Any future referral fee, sponsorship or marketplace commission must be displayed separately from PriceBolt’s ranking logic.