Reading Agent. Reads the inbound contract, identifies the document type and the controlling jurisdiction, and structures the clauses against the firm's clause taxonomy.

Playbook Agent. Compares each clause against the firm's playbook positions — fall-back, market-standard, walk-away — and tags every deviation with the relevant playbook reference.

Redline Agent. Drafts suggested redlines for deviations in the firm's preferred style; the associate or partner accepts, adjusts, or escalates per matter.

Risk Summary Agent. Produces a one-page risk summary for the partner covering material deviations, unusual terms and recommended client questions.

Search Agent. Translates the matter's legal question into structured queries across the firm's research subscriptions and open sources; collects relevant judgments and authorities.

Synthesis Agent. Reads collected authorities, identifies the controlling principles, distinguishing facts and reasoning, and assembles a synthesis with every assertion cited to the source paragraph.

Drafting Agent. Drafts a research memo in the firm's house style — issue, applicable law, analysis, conclusion — for the associate to refine.

Knowledge Agent. Files the completed research into the firm's matter knowledge base so the next associate working a related question starts from a stronger base.

Classification Agent. Reads each document, classifies relevance against the matter's issue tags, and produces a per-document confidence score; obvious cases go on a fast path, uncertain ones go to the review queue.

Privilege Agent. Identifies privilege candidates — attorney-client communication, work product — and surfaces them for confirmation with the supporting context.

Issue Tagging Agent. Tags documents by the issue or claim they relate to, so the review team can work by issue rather than by chronology.

Log Agent. Drafts privilege log entries from confirmed privileged documents, in the format the court or opposing counsel expects.

Intake Agent. Receives invoices across channels (email, portal, scan), classifies vendor and document type, and routes through the appropriate extraction template.

Extraction Agent. Pulls structured fields with per-field confidence — vendor, PO reference, line items, amounts, tax — surfacing low-confidence fields for human review rather than writing them silently.

Matching Agent. Three-way matches the invoice to the PO and goods receipt; auto-clears items within tolerance and routes mismatches to procurement with the discrepancy explained.

Approval Agent. Routes invoices for approval per policy (amount, cost centre, vendor), packages context for the approver, and posts to the ERP on approval.

Ingestion Agent. Pulls bank, credit-card and payment-processor transactions per client; identifies the vendor or counterparty even when descriptions are messy.

Categorization Agent. Categorizes each transaction against the client's chart of accounts using the client's history and the firm's playbook; surfaces low-confidence items for confirmation rather than guessing.

Anomaly Agent. Surfaces unusual amounts, new vendors, possible duplicates, and entries that look like personal-vs-business confusion, for the bookkeeper to investigate.

Reporting Agent. Prepares the monthly book-close summary per client with notes on items that needed judgment, so the partner reviewing has context not just numbers.

Intake Agent. Receives client documents (PBC items) across email and shared folders, classifies by type and audit area, and tracks the PBC list completeness against the audit program.

Evidence Agent. Reads documents and organizes them against the audit assertions they support — existence, completeness, valuation, presentation — with the relevant excerpts surfaced.

Work Paper Agent. Drafts work papers in the firm's house format with evidence references and tick-mark conventions; the auditor reviews and signs rather than originates.

Status Agent. Maintains a single audit status view across the engagement — what is complete, what is outstanding, what is pending review — so the partner sees the picture continuously rather than at status meetings.

Intent Agent. Reads session signals — browse path, search queries, abandoned cart, prior support conversations — and infers what the shopper is trying to do this session.

Merchandising Agent. Surfaces best-fit products with reasoning the shopper can see ("matches your saved size, similar fabric to last purchase"); refreshes recommendations as the session evolves.

Support Agent. Handles routine product-fit, sizing and policy questions end-to-end across web chat and WhatsApp; hands off the conversations that actually need a human.

Risk Agent. Scores refund and return probability at checkout against historical patterns and order signals; flags high-risk orders for ops review without blocking the storefront.

Enrichment Agent. Reads supplier-provided data — spec sheets, raw descriptions, images, product feeds — and produces a clean canonical product record with structured attributes.

Listing Agent. Generates marketplace-specific listings from the canonical record — Amazon-class title/bullets/A+ content, Flipkart-class category attributes, web storefront copy — respecting each channel's content rules.

Compliance Agent. Validates listings against marketplace policy before publish — prohibited terms, image requirements, category-specific attribute completeness; raises a checklist of fixes rather than silently failing.

Sync Agent. Keeps prices, stock, variants and content synchronized across channels; flags drift when a channel's view diverges from the canonical record.

Identification Agent. Identifies the customer and the order from the incoming message regardless of channel, even when the customer has not included an order number; pulls full order, shipment and payment context.

Policy Agent. Applies the returns and exchange policy to the specific case — window, condition, channel of purchase, item category — and identifies what resolutions are possible.

Resolution Agent. Where the case is clean, executes the resolution end-to-end — return label issued, exchange order created, refund initiated — and confirms with the customer.

Escalation Agent. For genuine exceptions (out-of-policy goodwill, damaged in transit, suspected fraud), packages full context and routes to a human agent.

Research Agent. Given a story brief, gathers public filings, prior coverage, regulatory documents, court filings and structured data sources relevant to the subject.

Summary Agent. Reads what was gathered and produces a structured background note — what is established, what is contested, what is unknown — with every assertion cited to a source.

Chronology Agent. Assembles a working timeline of events relevant to the story from the gathered materials; the reporter refines and uses it as a checking tool.

Knowledge Agent. Files the completed background into the newsroom knowledge base so the next reporter starting from a related angle does not start from zero.

Transcription Agent. Generates transcripts for the back-catalogue using the appropriate speech-to-text service per language and content type; aligns transcripts to timecodes.

Description Agent. Reads scenes (sampled frames + audio context) using commercial vision services and produces scene-level descriptions — what is happening, what is on screen, what is being discussed — with timecodes.

Tagging Agent. Identifies entities — people, places, organizations, topics — and tags timecoded segments; surfaces low-confidence identifications for archive-team confirmation.

Search Agent. Powers a search interface where producers ask in natural language ("a one-minute clip of the energy minister at the 2018 climate conference talking about renewables") and gets timecoded results.

Classification Agent. Classifies each comment against the publisher's policy taxonomy — toxic, spam, off-topic, ad hominem, factually contested, high-quality — with reasoning attached.

Action Agent. Auto-actions clear-cut cases per policy (hide spam, hold for review on borderline toxicity); never auto-removes content from the policy edge cases without human confirmation.

Response Agent. Drafts publisher-voice responses for comments the community team chooses to engage — corrections, clarifications, follow-ups — for the team to confirm before posting.

Insight Agent. Surfaces patterns across comments — what readers misunderstood, what corrections recur, what content provoked engagement — back to the editorial team.

Scheduling Agent. Books and rebalances appointments against real provider availability, patient preference and procedure-specific rules; handles cancellations and waitlist backfill.

Records Agent. Retrieves prior reports from connected systems and produces patient-readable summaries ahead of the visit, with full source citations for the clinician.

Preparation Agent. Generates procedure-specific prep instructions in the patient's language and drives reminder cadence over WhatsApp, SMS and email.

Consent Agent. Drives e-sign workflows for consent and intake documents, surfaces missing items, and keeps the EHR as the source of truth.

Records Retrieval Agent. Pulls the longitudinal record for the patient from FHIR-connected systems and the document store; identifies what is relevant for the upcoming visit context.

Summarization Agent. Produces a visit-ready summary — active problems, recent labs and imaging, medication list, recent events — with every clinical claim cited to a source document by reference.

Coding Support Agent.. Suggests ICD and procedure codes based on the documented encounter, with rationale cited to the note; the coder approves rather than originates.

Authorization Agent. Reads the procedure order, identifies the payer's specific pre-auth requirements, gathers the required documentation from the EHR, completes the payer's form, and submits — by API where the payer supports it, packaging documents ready for portal submission where not.

Documentation Agent. Surfaces missing-documentation requests from payers and pulls the required record fragments together for the clinician to confirm before submission.

Tracking Agent. Watches submitted authorizations and claims; pulls status; surfaces denials with reasons and recommended next steps; coordinates appeals where appropriate.

Knowledge Agent. Indexes course content — lecture videos with transcripts, slides, readings, prior Q&A — into a per-course knowledge base.

Answering Agent. Answers learner questions from the course materials with citations to the specific module and timestamp; admits when the question is outside the course scope rather than improvising.

Escalation Agent. Routes the questions the assistant cannot confidently answer to a human TA with the learner's context and what was attempted, so the TA picks up where the assistant left off.

Insight Agent. Aggregates the questions across learners and surfaces where the course content is unclear or missing, so the course team can improve materials cohort by cohort.

Engagement Agent. Responds to every inquiry within minutes on the channel it arrived, with knowledge of the specific programs, batches, fees and schedule.

Qualification Agent. Asks the prospect about their background, target outcome, timeline and city preference naturally inside the conversation; scores fit for the available programs.

Booking Agent. Offers counseling-session slots from the right counsellor's live calendar — the counsellor closest to the prospect's city or program — and confirms directly.

Nurture Agent. Runs adaptive follow-up for prospects who are not yet ready, with content matched to where they are in the decision.

Reading Agent. Reads each submission alongside the rubric and the assignment prompt; identifies which rubric criteria the submission addresses and how.

Scoring Agent. Drafts a score against the rubric with specific evidence cited from the submission; the instructor confirms or adjusts rather than starting from a blank score sheet.

Feedback Agent. Drafts personalized feedback aligned to the rubric and to where the student succeeded or fell short; the instructor edits and signs off.

Consistency Agent. Surfaces score-distribution patterns across the instructor team and flags places where grading is drifting, so consistency can be discussed before scores are released.

Engagement Agent. Responds to every inquiry within minutes on the channel it arrived — web chat, WhatsApp, email — with project-specific knowledge.

Qualification Agent. Asks budget, location and timeline naturally in the conversation; scores intent without making the prospect feel screened.

Booking Agent. Offers site-visit slots from the live sales calendar and confirms directly; handles reschedules and reminders.

Nurture Agent. Runs adaptive multi-channel follow-up for leads not yet ready, until they are or they clearly aren't.

Intake Agent. Accepts the property intake from the agent — basics, photos, raw notes — and produces a canonical property record with structured attributes.

Generation Agent. Generates listing copy from the canonical record — title, description, highlights, locality narrative — per portal's content rules and length limits.

Media Agent. Orders, captions and resizes photos per portal requirements; flags poor-quality images for re-shoot rather than publishing them.

Syndication Agent. Publishes the listing to each portal via the appropriate API or feed; keeps prices, status and content synchronized across portals; resolves drift when a portal's view diverges from the canonical record.

Intake Agent. Receives tenant requests across every channel; identifies the resident, the unit, the urgency, and the category (plumbing, electrical, appliance, common-area, lease question).

Triage Agent. Classifies the request — informational, scheduled maintenance, urgent — and either resolves end-to-end (lease questions, statement requests) or routes to the right path.

Update Agent. Keeps the tenant updated through the lifecycle and writes the activity to the owner portal so the owner sees the picture as it unfolds, not at month-end.

Intake Agent. Reads incoming tender documents — RFP narrative, BOQ, specifications, contractual conditions, technical schedules — and structures the requirements per discipline (civil, MEP, finishes, etc.) so each section reaches the right SME.

Compliance Agent. Builds the tender's compliance matrix from the document, identifies mandatory submissions, eligibility criteria and disqualifying clauses; flags anything the firm cannot meet so a no-bid decision can be made early rather than late.

Drafting Agent. Drafts technical and commercial response sections using the firm's prior winning proposals as the playbook — methodology, project experience, organizational details, similar-work references — for the proposal team to refine.

Coordination Agent. Tracks input requests across the response team, surfaces blocking items against the submission deadline, and assembles the final submission pack against the tender's submission checklist.

Classification Agent. Reads incoming submittals and RFIs; identifies the discipline, the specification section referenced and the reviewing consultant from the transmittal cover and document metadata; tags by criticality against the construction schedule.

Routing Agent. Routes each item to the right consultant or internal reviewer with the relevant specification clauses and prior responses already attached — the reviewer does not search for context, the context arrives with the item.

Tracking Agent. Watches review deadlines against contractual response windows; surfaces overdue items to project management with the chain of communication already documented; flags items where the response is inconsistent with prior responses on the same project.

Coordination Agent. Maintains the single shared view of every open submittal and RFI across the project — what is in review, what is approved, what is rejected, what is overdue — so the weekly project review reads a current ledger rather than reconstructing one.

Ingestion Agent. Reads daily progress reports filed across formats (PDF, Excel, photo-attached emails, app uploads) and extracts structured data — activities completed, manpower deployed, equipment used, materials consumed, safety incidents, weather conditions — and writes to a per-project database.

Photo Agent. Organizes site photos against the day's activities and locations using captions and EXIF data; surfaces a visual progress timeline per work package without anyone curating it by hand.

Schedule Agent. Cross-checks reported progress against the master schedule (MS Project / Primavera export); flags activities running behind, ahead or off-sequence; produces a current schedule-variance picture for project review.

Insight Agent. Aggregates patterns across daily reports per project and across the portfolio — recurring delays, manpower shortfalls, materials always running late, safety near-misses concentrated on specific activities — so leadership sees signal, not noise.

Conversational Agent. Takes orders over WhatsApp and voice in the customer's language; suggests sensible upsells based on order history; confirms specifics before sending to the kitchen.

Orchestration Agent. Normalizes orders from every channel into a single kitchen queue; resolves modifier and item conflicts before they hit the line.

Capacity Agent. Monitors prep load in real time and adjusts promised times honestly during peaks; surfaces a degraded state instead of accepting orders the kitchen cannot deliver.

Dispatch Agent. Batches and assigns delivery riders based on prep state and proximity, so food leaves the kitchen at the right moment.

Listening Agent. Continuously ingests reviews and ratings across platforms per outlet; classifies sentiment, topic and the underlying operational dimension (food, service, delivery, packaging, value).

Response Agent. Drafts brand-voice responses to negative and notable reviews, with the marketing team approving before public posting; never replies to unverified-fact claims without escalation.

Pattern Agent. Aggregates signals per outlet, per dish, per time-of-day and per delivery partner to surface where the operational problem actually lives; feeds findings to the operations team.

Recovery Agent. Where appropriate and within policy, initiates a goodwill recovery — a credit, a re-delivery, an apology — for affected customers, with human approval for anything above a threshold.

Suggestion Agent. Generates short-horizon demand suggestions per kitchen and per dish using recent order history, day-of-week patterns, menu changes, weather and known local events as inputs the chef can weight or override.

Prep Agent. Translates the chef's confirmed plan into prep lists per shift, respecting prep rhythm and shelf-life constraints; surfaces a recommended prep plan the head chef confirms.

Ordering Agent. Suggests ingredient reorders against the confirmed plan plus current stock, with lead times for each supplier; flags items where the demand picture is uncertain so the chef decides rather than the system over-ordering on its behalf.

Waste Agent. Captures actual waste as it is recorded and feeds it back into the daily suggestion view, so the patterns surface within days, not at month-end.

Onboarding Agent. Gathers and validates supplier documents, runs structured background checks, and coordinates the parallel reviews across procurement, compliance and finance.

Monitoring Agent. Tracks each active supplier against news, sanctions and regulatory feeds, ESG signals, and delivery-performance history from the ERP; raises a flag the moment something changes materially.

Scoring Agent. Maintains a continuously updated supplier profile and risk score, with rationale, so the score is defensible — not a black box.

Reporting Agent. Drafts procurement memos comparing alternative suppliers when a sourcing decision comes up, with current data attached.

Signal Agent. Watches time-series sensor data per machine against configured thresholds and known failure signatures; flags anomalies for engineer attention with the raw signal context attached.

Knowledge Agent. When a signal fires, retrieves the relevant equipment manual section, the maintenance history for that asset, and the standard procedure for the suspected fault — assembled into a single ready-to-act briefing.

Work Order Agent. Drafts a work order with the suggested fault, the parts needed, the procedure to follow, and the priority based on production schedule; the maintenance manager approves and dispatches.

History Agent. Captures the actual resolution back into the knowledge base, so the next time the same signature appears the briefing is sharper.

Document Agent. Reads engineering drawings, datasheets, change notices and BOM exports as they enter the document management system; structures the content for downstream validation.

Validation Agent. Cross-checks the change against the existing BOM, the parts catalog, applicable standards, and tolerance rules; flags inconsistencies — orphan parts, missing references, incompatible variants — with traceable rationale.

Notification Agent. Routes findings to the right team — engineering, procurement, quality — with the underlying document context attached, so the recipient does not have to search for what changed.

Vision Agent. Reads existing camera feeds via commercial computer-vision services; detects vessels at the harbor entrance and surfaces them with timestamped evidence for the operations team.

Document Agent. OCRs and structures bills of lading, customs paperwork and manifests on arrival; classifies and routes them through the right verification flow.

Reconciliation Agent. Cross-checks vessel observations, manifest data and container-side readings; flags inconsistencies for human review before clearance.

Workflow Agent. Drives the verification checklist for berth allocation and clearance, keeps every party informed, and logs every decision.

Ingestion Agent. Receives documents per shipment across channels (email, EDI, portal upload), classifies by document type and trade context, and assembles the per-shipment document set.

Extraction Agent. Pulls structured fields with per-field confidence — HS codes, declared values, parties, weights, dimensions, regulatory references — surfacing low-confidence fields for human review.

Filing Agent. Drafts the customs declaration or pre-filing for the destination using the extracted fields, applies destination-specific rules, and presents a ready-to-submit filing to the licensed broker.

Validation Agent. Cross-checks documents against each other for internal consistency and against destination-country requirements before filing.

Picking Agent. Builds picking sequences per order or wave that respect the actual layout and the day's order profile; configurable rules update as the floor manager refines them.

Replenishment Agent. Watches forward-bin levels against the day's pick demand and triggers replenishment ahead of the line going dry, not after.

Inventory Reconciliation Agent. Detects inventory discrepancies as they appear in pick exceptions — short picks, item-not-found, wrong item — and reconciles or escalates rather than letting them roll into the next cycle count.

Coordination Agent. Coordinates picking, replenishment and inventory across the shift so the three run to a shared picture, not three separate ones.