Glossary

Activity feed — the real-time log panel that shows what an AI agent is doing during a run. Displays actions like reading documents, running searches, evaluating criteria, and recording findings as they happen.

AI provider — the AI model provider used for analysis. You can choose between Anthropic (Claude), OpenAI (GPT), and Google (Gemini). The system automatically selects the right model tier for each task.

Analysis — the AI-powered process of comparing a vendor's proposal against your procurement requirements. Produces findings, scores, and recommendations.

Analysis run — a single execution of analysis on one bid. Each time you click "Analyze", that's one run.

Autopilot mode — a composition / protocol mode where the agent runs straight through to completion without pausing for review. Fastest, but you only see the result at the end.

Batch analysis — analyzing multiple bids in sequence with a single click ("Analyze All").

Bid — a vendor's submission for your procurement. Contains one or more uploaded documents (the proposal, supporting materials, etc.).

Bid composition — drafting a vendor's bid documents from a planned outline. Goes through a setup interview, an approved compose plan, the composition itself, and a validation pass. See Composing a bid.

Captured entity — a person, company, or organization mentioned in the analyzed documents. Automatically detected and, for Latvian procurements, matched against the official business register.

Clarification request — a question the AI asks you during a run when it encounters something ambiguous. The run pauses (with its workspace preserved) until you respond.

Compose plan — the outline produced by a setup interview before composition starts. Lists the artifacts (sections / documents) to be drafted, the mode, tone, output formats, and library attachments. You review and approve the plan before composition runs.

Compose plan artifact — one planned deliverable inside a compose plan. Has a status (planned → drafted → published, or needs_review / skipped) and links to the published document once drafted.

Compose validation — a two-pass review that runs automatically after bid composition completes. Pass 1 (coverage) checks structural gaps against the RFP; pass 2 (content) flags filler text, unsubstantiated claims, and inconsistencies. See Compose validation.

Coverage gap — a missing or incomplete section flagged by the validator's coverage pass. Each gap cites the specific RFP excerpt that wasn't addressed.

Credits — the currency used to pay for AI runs on the platform. Purchased via Stripe at 6 EUR per credit.

Critical — the highest severity level for findings. Indicates a potential deal-breaker, such as failing a mandatory requirement.

EDOC — a digitally signed document archive format used in Latvia. The platform extracts and parses all files inside automatically.

Evaluation criteria — scoring rules you define for a procurement. Either pass/fail or scored (0–100). The AI extracts values from each proposal based on these criteria.

Evaluation protocol — the post-evaluation procurement document required by Latvian law. The platform's protocol agent drafts it from analysis results and the RFP. See Evaluation protocol.

Evidence — exact quotes from documents that support a finding. Included with every finding so you can verify the AI's reasoning.

Finding — an issue or positive note flagged by the AI during a run. Has a severity level (Critical, Major, Minor, or Strength) and a provenance badge showing which agent identified it.

Finding category — the area a finding relates to. In bid analysis, categories are tied to evaluation criteria. In RFP quality review, the eight categories are: Clarity, Completeness, Consistency, Fairness, Criteria Quality, Timeline, Legal, and Budget.

Funding research — an AI feature that searches the web for relevant funding opportunities (grants, calls, programmes) at the organization level. See Funding research.

Guided mode — a composition / protocol mode where the agent pauses after each section (or whenever it has something for you to review). You give feedback and click Continue. Slower than autopilot but gives you control.

Interview mode — a composition mode where the agent asks you questions throughout the drafting process, in addition to pausing for review. Best for complex bids with high-touch sections.

Library — the organization-level repository of reusable vendor documents (CVs, certifications, past projects, etc.). The bid composition agent can pull from the library automatically when sections need supporting evidence.

Lot — a sub-item within a procurement that vendors can bid on independently. Common in large public procurements.

Major — the second-highest severity level. A significant gap or unclear response that's risky but not necessarily a deal-breaker.

Market research — an AI feature that searches the web and compiles structured findings about products, suppliers, regulations, and risks. See Market research.

Minor — a low-severity finding. Small omissions or formatting issues unlikely to affect the overall outcome.

Orchestrator — the main AI agent that drives a run. Reads documents, plans what to check, calls tools, evaluates evidence, and decides when the work is done.

Parsing — converting uploaded documents (PDF, Word, etc.) from their visual format into structured Markdown the AI can read and search. See Document parsing.

Pause / resume — when an agent needs your input, it pauses, preserves its sandbox workspace, and waits. After you respond, it resumes from exactly where it stopped. No lost progress, no time limit.

Preparation checklist — an AI-generated list of items to prepare before submitting a bid. Items are organized by category and priority level (required, recommended, optional). Generating a checklist is free.

Procurement — your project workspace for one purchasing process. Contains RFP documents, vendor bids, analysis results, and comparison data.

Recommendation (RFP review) — the overall assessment at the end of an RFP quality review. One of four levels: Ready to Publish, Minor Improvements Needed, Significant Revision Needed, or Not Ready for Publication.

Revision round — when validator findings need work, you can apply your feedback and trigger one more compose pass. Currently capped at one revision round per bid composition.

RFP — Request for Proposal. The documents describing what you're looking for — requirements, specifications, evaluation rules, contract terms.

RFP composition — drafting the RFP document set on the buyer side. The agent produces structured sections you can edit. See Composing an RFP.

RFP integrity check — an AI scan that flags fairness and integrity risks in the RFP, such as vendor-specific language or overly narrow specifications. See RFP integrity check.

RFP quality review — an AI-powered review of your RFP documents for quality, clarity, completeness, and compliance before publishing. Produces findings with severity levels across eight categories and an overall recommendation. Costs credits.

Sandbox — an isolated workspace (a fresh, throwaway environment) created for a single run. The agent uploads documents into it, reads them, makes its decisions, and the sandbox is destroyed when the run finishes. Provides privacy isolation between runs and across organizations.

Semantic search — a fallback search method the agent uses when keyword search can't find a fuzzy concept. A smaller AI model reads the relevant files and ranks passages by meaning.

Setup interview — a short interview the bid composition flow runs before drafting starts. Produces a compose plan for you to review and approve. See Setup interview.

Signature detection — automatic detection of digital signatures on uploaded documents. Documents with detected signatures show a signature badge.

Specialist — a focused, single-purpose agent run that prepares the ground for the main analysis. Bid analysis runs two specialists before the main pass: one extracts evaluation criteria from the RFP, one extracts lots.

Strategic briefing — the executive summary at the top of analysis results. Includes a recommendation (Strong Submit, Submit with Improvements, Significant Revision Needed, or Do Not Submit), finding counts, and a plain-language summary.

Strength — a positive finding. Something the vendor did well or exceeded requirements on.

Submittable document — a prep-checklist item that requires an actual file (e.g. a CV, a certificate). The bid composition flow uses these to plan which documents need to be attached or generated.

Verification — a quality control step in bid analysis where a separate AI agent reviews Critical and Major findings. Can confirm, downgrade, or remove findings.

Verification coverage — a status indicator showing what fraction of a run's serious findings have been double-checked by the verification phase.