FAQ
Pricing, how Benchside compares, how it handles AI vendors, and how your data stays yours.
It's pay-as-you-go, per project. Your first project is free and no payment method is needed to begin, so you can ship a full intelligence package before you ever enter a card. Both modules - Scope and Evaluate - are included in every project, so you pay for work done, not for which capability you unlock. Enterprise adds custom contracts, white-label exports, audit-log export, and an SLA.
Both. The engine detects the deal scale and right-sizes everything. A 24-person startup evaluating a $40K SaaS tool gets the 10 to 15 questions that matter, dollar figures sized to that deal, and the SaaS-specific traps (auto-renewal, the SSO tax, data export on exit). A Fortune 500 running a $40M programme gets the full enterprise depth. Same frameworks, fitted to you.
Vendr helps you negotiate a better price after you've decided to buy. Zip manages the internal approval workflow for your purchase. Both activate after the vendor decision is made. Benchside activates before - it generates the interrogation kit, scope package, and architecture map you need to evaluate whether the vendor is right and what risks you're taking on. Use Benchside first. Use Vendr or Zip after.
A team of agents, each owning a stage. A lead agent (Deal Monitor) assesses the deal and dispatches specialists: scope authoring, scope-drift detection, a commitment ledger, gap interrogation, a demo-to-reality auditor, and a live session co-pilot. They reason across steps, check their own work, and get sharper on a vendor with every deal you run. Ben, your copilot, can answer across all of it.
Generic tools give you a checklist. Benchside applies named, source-backed frameworks against a deep enterprise knowledge base calibrated to your specific project type, vendor, budget, and tech stack before generating a single output. You get a scope package with dollar-estimated exclusions, architecture decisions with lock-in rankings, and interrogation questions with good/red-flag answer guides. The output is structured and typed, not a chat response.
Yes, and it is the category where the asymmetry is widest right now. Benchside profiles OpenAI, Anthropic, Azure OpenAI, AWS Bedrock, and Google Vertex AI, and generates the AI-specific questions and clauses generic procurement frameworks miss: contractual model-deprecation and behaviour-change notice, training-data indemnification scope, a quality SLA distinct from uptime, agentic spend ceilings with auto-pause, fine-tuned weight return on exit, plus EU AI Act and ISO 42001 coverage. Select the AI / ML Platform project type, or just name the vendor, and the kit calibrates to it.
No. Your project data is organisation-isolated behind row-level security. No data is shared across tenants. The vendor has no visibility into your platform. Even Benchside staff cannot query your project data.
No. Your project description, vendor context, and generated outputs are used only for the generation call, never retained or used for model training. We operate under a strict data processing agreement with our AI infrastructure providers that prohibits training use.
Yes. Pay-as-you-go includes role-based access (admin, member, viewer) and two-factor authentication for your whole team. Enterprise adds SCIM provisioning and unlimited seats, with SAML SSO on our near-term roadmap.
Start your first project free, or talk to us - no payment method needed to begin.