Semarize
Semarize is a conversational intelligence API that turns calls, emails, chats and transcripts into typed fields, scores and risk flags, each carrying a confidence figure and cited evidence, for revenue, data and quality teams that already own a recorder.
What is Semarize?
Semarize describes itself as The Conversational Intelligence API: a programmable semantic layer rather than an analytics platform. It does not record, does not transcribe and does not send a bot into meetings. It sits behind the capture tools a team already runs — Gong, Chorus, Zoom, Microsoft Teams, Google Meet, Cisco Webex, Aircall, Clari, Fathom, Fireflies, Granola, Otter, Outreach, Salesloft — and turns what they produce into structured data a CRM, a BI tool or an internal application can act on, without anyone reading the transcript.
The product rests on three primitives. A Brick is one reusable semantic check about a conversation: a single question returning a typed value — boolean, numeric score, category or list of strings — together with a confidence figure, a reason, and the exact transcript span that justifies the answer. Bricks are versioned, so changing a scoring rubric creates a new version and historical runs stay attached to the logic that produced them. A Kit groups Bricks into one evaluation framework executed together, identified by a kit_code. A Run is one conversation evaluated against one Kit.
Evaluation is anchored by knowledge grounding: the customer uploads product documents, playbooks, price lists and policies into a knowledge base, and Bricks then assess conversations against that real business context, which is how outdated pricing or stale messaging gets flagged rather than merely keyword-matched. Accepted inputs include sales calls, support chats, internal meetings, recorded video, transcripts and email threads — anything that arrives as audio, transcript, or text.
Access is developer-facing. A POST to /v1/runs with a kit_code and an input returns 202 asynchronously, with results delivered by HMAC-signed webhook. A JSON-RPC 2.0 MCP server exposes eleven tools and offers OAuth connectors for Claude and ChatGPT alongside session-based machine agents. Seven team pages and seven industry pages all rest on the same Bricks, Kits and API foundation.
The publisher is Semarize Ltd, a British company incorporated on 26 March 2026 and founded by Alex Handsaker after a career in technology sales and enablement. Its stated audience is RevOps teams, data teams, SaaS operators, GTM engineers and builders, and it presents itself as programmable semantic layers connecting conversations to operational systems. The claimed limitation is deliberate: no transcription, no new bots, and none of the platform baggage.
What it does
- Extract CRM fields from conversations: budget, timeline, stakeholders, competitors, agreed next step.
- Score every call against an internal methodology — MEDDIC, MEDDPICC, BANT or a bespoke framework.
- Raise risk flags: qualification drift, invalid next step, pricing-policy breach, outdated messaging.
- Run automated QA on 100% of interactions, covering script adherence, regulatory compliance and resolution quality.
- Detect churn risk, expansion intent and renewal signals inside customer conversations.
- Evaluate AI agents at scale: chatbot quality, hallucinations, adherence to instructions.
- Feed a data warehouse with typed conversational facts, and structure recruitment interviews with competency scoring and bias checks.
When to use Semarize / When not to
A quick filter to help you decide if Semarize is the right fit.
When to use Semarize
- Sales teams and leaders who already record calls and want budget, timeline, stakeholders, competitors and the agreed next step written back into CRM fields as typed values.
- Sales enablement and coaching teams measuring the before-and-after impact of a programme, rep by rep, against MEDDIC, MEDDPICC, BANT or an in-house discovery framework.
- Revenue operations leaders and BI or data teams feeding a warehouse with typed conversational facts instead of free-text summaries, then reporting on them.
- Customer success, quality assurance and compliance teams: churn and renewal signals, QA across 100% of interactions, and regulatory checks with an evidence trail, including contact centres, financial services, insurance and interview scoring in recruitment.
- AI engineers, product teams and GTM builders evaluating AI agents at scale and using Semarize as the structured data layer underneath internal tools built with AI coding assistants such as Claude and Codex.
When not to use Semarize
- Teams looking for a transcription or recording tool: Semarize does not transcribe, it consumes transcripts produced by others and sits behind the tools already in place.
- Anyone wanting a note-taker or a meeting bot — the pitch is explicitly no new bots and no rip-and-replace, and nothing dials into calls.
- Managers wanting a ready-made analytics platform with a review interface: the product is sold without the platform baggage and offers no dashboard front end of its own.
- Non-technical buyers expecting one-click native connectors or a mobile app: there is no iOS or Android application, ingestion runs through the API or a third-party automation platform, and an API key, a Kit and a routing pipeline are needed.
- Organisations wanting to automate employment, legal, credit, insurance, medical or immigration decisions: clause 6.8 of the terms forbids using the probabilistic outputs as the sole basis for any of them.
How to use Semarize
A typical end-to-end flow, from setup to results.
- Create an account at app.semarize.com/signup — no card required, with 250 starter credits.
- Define your Bricks: name the signal (for example next_step_confirmed), choose the output type (boolean, score, category or list), write the rubric, and require evidence and confidence.
- Group the Bricks into a Kit per workflow or per team; the Kit receives a kit_code.
- Optionally attach a knowledge base — product docs, playbooks, price lists, policies — so evaluation is grounded in your own documents, at +5 credits per run.
- Generate an API key: smz_test_* for the sandbox, smz_live_* for production. The key is displayed only once, at creation.
- Bring conversations in through the API, through Zapier, Make, n8n, Clay or Workato, or by following the published per-platform extraction guides.
- Call POST /v1/runs with the kit_code and the input; the response is an asynchronous 202 queued, and an optional Idempotency-Key header makes retries safe.
- Collect the result through the HMAC-signed webhook or by polling, deduplicating on the Semarize-Delivery-Id header.
- Route the typed JSON to your CRM, warehouse, BI tools, Slack alerts or internal applications.
- Alternatively, connect the MCP server to Claude or ChatGPT to compose and publish Bricks and Kits without writing application code.
Pros & Cons
Pros
- Typed, deterministic output that drops straight into a CRM or warehouse column, with no parsing needed.
- Evidence cited on every signal, so a human can audit the answer, and versioned Bricks that keep historical runs comparable after a rubric change.
- Evaluation grounded in the customer's own documents — pricing, playbooks, policies — rather than on keyword matching.
- No migration: it plugs in behind the recording tool already in place, with no extra bot dialling into calls.
- Readable pricing tied to actual work: 1 credit equals 1 Brick evaluation, bricks multiplied by conversations, on top of a permanent free tier of 250 credits requiring no card.
- Data ownership asserted contractually, derived outputs included, with an explicit commitment not to train general-purpose models on identifiable customer data.
- Broad integration surface (25 extraction guides, 5 automation platforms, 5 CRMs, 8 data destinations), an official MCP server for Claude and ChatGPT, and a fully published publisher identity: company number, registered office, incorporation date.
Cons
- No security certification is claimed — no SOC 2, no ISO 27001, no HIPAA. The only mention is prospective: certification standards the company may obtain from time to time.
- A very young publisher: incorporated on 26 March 2026, a single director, and a first accounting period not yet closed, so no verifiable operating history.
- No data hosting location is published — no country, no region, no named provider — and the subprocessor list is disclosed by category only, available upon request.
- No documented opt-out from training, despite the broad licence in clause 3.2(b) over aggregated and anonymised data, and no published list of supported languages.
- No contact page and no demo page: every Contact us and Talk to sales button leads to the signup form.
- No mobile app and no one-click native connector: ingestion goes through the API or a third-party automation platform.
- Commercial fine print: fees are non-refundable and unused credits are lost, billing disputes lapse after 14 days, unlimited kits on Platform and Enterprise carries a Fair Use asterisk, prices exclude VAT and the paid entry point is in pounds sterling, and beta features come with no warranty or SLA and can be withdrawn without notice.
Pricing & Plans
Semarize offers a permanent free plan. The Builder tier, billed pay-as-you-go, costs GBP 0 per month and includes 250 starter credits, 5 active kits, sandbox and API access, with no card required. The lowest paid entry point is the Team (Plus) plan at GBP 70 per month, which includes 1,500 credits, 25 active kits and 5 knowledge bases; Platform (Pro) is GBP 200 per month and Enterprise is quoted on request. Consumption is measured in credits, where one credit equals one Brick evaluation: a Kit of 8 Bricks therefore consumes 8 credits per run, with +5 credits when a knowledge base is attached and +1 credit for Reasoning & Proof. Published examples include a 6-brick discovery quality Kit over 100 conversations at 600 credits, a 7-brick MEDDICC scorecard at 700 credits and a 12-brick QA compliance Kit at 1,200 credits. Additional credits are sold at GBP 15 for 500, on a slider from 500 to 5,000. Amounts are displayed in pounds sterling by default with a USD toggle, and exclude VAT or sales tax. Fees are non-refundable, including unused credits.
- 250 starter credits
- 5 active kits
- sandbox
- API access.
- 1
- 500 credits included
- 25 active kits
- 5 knowledge bases
- workspace roles and permissions
- email support
- sandbox
- API access.
- 5
- 000 credits included
- unlimited kits (subject to the Fair Use Policy)
- 20 knowledge bases
- roles and permissions
- priority support
- sandbox
- API access.
- bespoke credit allocation
- unlimited kits and unlimited knowledge bases (subject to the Fair Use Policy)
- SSO and advanced access control
- dedicated SLA
- dedicated success manager
- custom retention.
- Additional credits — GBP 15 for 500 credits
- available on top of any plan.
Data, GDPR & hosting
A consolidated view of how Semarize handles your data.
GDPR overview
GDPR implementation is documented in detail. Semarize Ltd (company number 17118292, 128 City Road, London EC1V 2NX) declares a dual role: controller for website visitors, account holders and business contacts, processor for personal data customers submit to the platform. Legal bases under the UK GDPR are contractual necessity, legitimate interests, consent and legal obligation. Clause 8 of the terms sets out full processor obligations: breach notification without undue delay, an annual audit right at the customer's expense, and no transfer outside the UK or EEA without adequate safeguards and customer consent, subject to an explicit reservation for infrastructure and AI model providers. Data subject rights — access, rectification, erasure, restriction, objection, portability — are exercised at privacy@semarize.com, with a right to complain to the ICO. No Article 27 EU representative and no DPO are named. Both policies are dated 26 March 2026.
Who owns the data?
Clause 3.3(a) of the terms states that the customer retains all rights, title, interest and intellectual property in Your Data — transcripts, recordings, metadata, documents, configuration inputs and prompts. Derived results, including structured outputs, semantic scores, evaluations and reports, are contractually part of Your Data between the parties. Semarize keeps the intellectual property in its platform, models, evaluation methodologies, API and architecture. Clause 3.2 grants Semarize a non-exclusive, sub-licensable licence to use Your Data to perform the contract and, in aggregated and anonymised form, for performance reporting, service improvement and developing new products for the customer and other customers. The privacy policy makes the customer controller, Semarize processor, and states personal data is not sold.
Reuse rights
Yes. Because structured outputs count as Your Data, the customer can reuse and route them without asking permission — into CRM fields, a data warehouse, BI dashboards, alerts or internal applications — and the site makes portable, customer-owned typed output an explicit argument against platforms that keep the data captive. The restrictions in clause 3.4 bear on the platform, not on the results: no reselling or redistributing the service, no reverse engineering of the models or API, no derivative works of the platform, no removal of proprietary notices. The acceptable use policy adds a ban on extracting system prompts, scoring logic or evaluation methodologies. On exit, clause 8.5(g) provides for deletion or return of personal data on written instruction after termination, while clause 8.8 places the burden of exporting data before termination on the customer; after a free trial that is not continued, submitted data may be irretrievably deleted.
Data retention & training
Hosting summary
No hosting country, region or provider is named anywhere on the site. The security page says only Secure cloud infrastructure: conversation content is processed transiently and not retained beyond the customer's configured retention settings, while run metadata is stored so it stays queryable. Encryption is TLS 1.2+ in transit and at rest, and API keys are stored hashed, the raw key never being kept. Workspaces are partitioned, so one workspace's data is not reachable from another through the public API. On jurisdiction, clause 8.5(d) undertakes not to transfer personal data outside the UK or EEA without adequate safeguards and the customer's consent, with an explicit reservation for infrastructure and AI model providers; the privacy policy frames such transfers with UK-approved standard contractual clauses, adequacy decisions or equivalent safeguards. Subprocessor categories are published — cloud hosting, authentication, payment, analytics, AI model providers — but names are given only on request. For information, DNS resolution returned 216.198.79.1, a United States anycast node on Amazon AS16509: a technical observation, not a statement by the publisher.
Where Semarize works
Country-level availability.
Not available in
Things to keep in mind
Risks and trade-offs to weigh before adopting Semarize.
- Semarize processes recordings and transcripts, and therefore personal data belonging to third parties — customers, candidates, employees — who are not its users. Clause 8.3 places the entire burden of obtaining consents and legal bases on the customer, and the acceptable use policy forbids submitting data collected in breach of call-recording law. Scoring 100% of interactions also means staff and candidates are evaluated continuously.
- Clause 6.8 states that outputs are probabilistic and must not be the sole basis for employment, legal, credit, insurance, medical or immigration decisions, while the Recruitment and Financial Services pages target exactly those settings. Human review is a contractual obligation, and the temptation to treat a confident-looking score as a verdict is precisely the failure mode the clause guards against.
- No security certification is claimed — no SOC 2, ISO 27001 or HIPAA — and the only occurrence of the word is prospective. The publisher was incorporated on 26 March 2026, with one director who is also the sole person with significant control, and first accounts are not due until 26 December 2027.
- The hosting location is never stated, and the terms explicitly reserve the use of infrastructure and AI model providers implying international transfers. The subprocessor list is available upon request only, with categories alone published. No Article 27 EU representative is designated, so for a European controller the only channel is privacy@semarize.com in the United Kingdom, and no minimum age is stated anywhere.
- Clause 3.2(b) grants a broad, sub-licensable licence to reuse data for developing new and different products for the customer and for other customers, conditioned on aggregation and anonymisation. No opt-out mechanism from training is documented. The commitment not to train covers identifiable data and general-purpose models; it does not exclude aggregated use to improve the service.
- Commercial asymmetries worth reading before signing: fees are non-refundable including unused credits, a billing dispute is time-barred after 14 days, a chargeback is treated as a material breach, Semarize's own logs are the sole billing record, unlimited kits carries an unquantified Fair Use asterisk, and beta features come with no warranty or SLA and may be withdrawn without notice.
- Positioning to read carefully: the homepage FAQ says Semarize does not replace Gong, Chorus, Fathom or Otter, while five blog posts are titled after those same tools as alternatives. There is no contact or demo page — Contact us and Talk to sales lead to the signup form, so there is no human interlocutor before an account is opened. Pricing is shown in GBP by default with a USD toggle, so the figure depends on the display.
Setup & Integrations
Technical difficulty
Moderate to high, developer-oriented. The standard path is a Bearer-authenticated POST /v1/runs carrying a kit_code and a transcript, plus a webhook endpoint with HMAC signature verification. Simpler routes exist: Zapier, Make, n8n, Clay and Workato are documented, and the MCP server lets Bricks and Kits be composed from Claude or ChatGPT. The real effort is design rather than plumbing — defining each Brick's signal, type, rubric and evidence requirement, then assembling Kits. No data migration is needed, a sandbox with test keys is available on every plan, and 25 extraction guides cover getting transcripts out of existing platforms.
Deployment
Integrations
Behind Semarize
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Semarize.
Frequently asked questions
What is Semarize?
Does Semarize replace our meeting recorder or call platform?
What kinds of conversations can it process?
What exactly does it return?
Can we score conversations against our own framework?
Does it understand our product and our pricing?
How much does it cost?
Is there an API and an MCP server?
Who owns the data and the outputs?
Is customer data used to train models?
Should you pick Semarize?
Semarize is a narrow product, and deliberately so: a structuring layer, not a conversational analytics suite. What it does well is visible on its own pages. Outputs are auditable because every signal carries a confidence figure, a stated reason and the exact transcript span behind it. Bricks are versioned, so a rubric change does not silently invalidate past runs. Evaluation is grounded in the customer's own pricing, playbooks and policies. Data ownership is asserted in the contract, derived outputs included. And credit pricing has arithmetic — bricks multiplied by conversations — that can be worked out before signing.
The reservations are equally documented. The publisher was incorporated on 26 March 2026, has a single director and no closed accounting period. No security certification is claimed, no hosting country, region or provider is named, and subprocessors are disclosed by category only. No retention duration is quantified in any of the five legal documents. Anyone buying for a regulated environment should note that clause 6.8 of the terms forbids using probabilistic outputs as the sole basis for employment, credit, insurance, legal, medical or immigration decisions, while the recruitment and financial services pages address precisely those contexts: human review is a contractual requirement the marketing does not restate.
Practically, the tool assumes two things — a source of transcripts already in place, and enough technical means to call a REST API or wire up an automation platform. Given those, the permanent free tier of 250 credits, five kits and a sandbox, with no card required, makes evaluation genuinely low-cost. The questions worth putting to the vendor before committing are equally clear: where the data is hosted, who the subprocessors are, what retention windows apply, and what the Fair Use thresholds behind unlimited kits actually amount to.
- Choosing a selection results in a full page refresh.
- Opens in a new window.