Cresta Reviewed: Strong Integrations, Unpublished Everything Else
The integration list is the best-evidenced thing on the entry; pricing, deployment options and compliance frameworks are not published at all, and the review says so.
Reviewing Cresta — profile, published facts, and sources.
In short
What Cresta does well
- Publishes a specific, checkable integration list spanning telephony, CRM, knowledge and data systems, rather than a generic claim of connectivity.
- Evaluation, simulation and regression testing are described as part of the platform rather than as a practice the customer has to build.
- Customer deployments are published with the customer named, not anonymised.
Where it falls short
- Publishes no pricing of any kind, so cost cannot be estimated without contacting sales.
- No on-premise or private-cloud deployment option is described in public material; the platform is presented as cloud only.
- Compliance badges appear without naming the frameworks they represent, so certifications cannot be verified from published material.
What it is
Cresta sells an AI platform for contact-centre conversations. It runs in two modes that share the same underlying models: an autonomous agent that handles voice and digital conversations end to end, and an assistive layer that coaches human representatives in real time while they are on the call. Around both sits conversation analysis and automated quality management.
It is layered onto an existing contact-centre stack rather than replacing it. That is the architectural decision that explains most of the ratings below: it makes integration breadth the product’s central problem, and it means the platform inherits whatever deployment and compliance posture the underlying stack already has.
Where it is strong
The integrations page names its systems. That sounds like a low bar and it is not one most of this market clears: telephony and CCaaS platforms, CRM and case management, knowledge sources, and the data warehouse are each listed by product name. A buyer can hold that list against their own stack in about five minutes, which is the only useful test of an integration claim.
Evaluation is treated as platform functionality rather than a customer exercise. The overview describes LLM-based evaluation, simulation against prior conversations, and regression testing before release, with a separate quality management product reviewing live conversations afterwards. The before-release and after-release halves are both present, which is unusual.
Customer deployments are published with the customer named. An anonymised case study is an assertion; a named one is something a prospective buyer can go and ask about.
Where it falls short
There is no published pricing of any kind — no rate card, no starting tier, no unit of consumption. Every path on the site ends at a demo request. For a buyer trying to build a shortlist, that is a hard stop rather than an inconvenience, and it is the single largest gap on this entry.
Public material describes one deployment model: a cloud platform. No private-cloud, on-premise, data-residency or customer-managed-key option is described anywhere. Whether such options exist under NDA is exactly the point — this review can only assess what is published.
Compliance badges appear on the site without naming the frameworks behind them. A badge that does not say SOC 2 Type II, or ISO 27001, or HIPAA, and does not link to a report or an audit date, cannot be verified by the reader and is not treated here as evidence.
None of the capability claims carry a number. There is no published containment rate, resolution rate, or accuracy figure with a methodology attached, so the size of the end-to-end claim cannot be checked. That is why agent capability sits at four rather than five below, despite the product being built for it.
Who it suits
A contact centre with an existing CCaaS and CRM investment it is not planning to replace, enough conversation volume to justify a sales-led purchase, and a procurement process that can extract deployment and compliance detail privately. Buyers who need to price a project from public material, or who need a documented residency story before the first meeting, will find the entry short of what they need.
What would change our mind
Published pricing, named compliance frameworks with dates, or a single containment figure with its methodology would each move a rating below. None of them requires the product to change — only the public material.
Ratings
These scores exist to decide the awards, and that is all they are for. The assessment is the article above; this is the arithmetic it feeds. Each criterion is scored 1–5 and weighted per the published rubric (v0.1-draft).
| Criterion | Weight | Score |
|---|---|---|
| Agent capabilityPublishes an autonomous voice and digital agent that handles conversations end to end, alongside assistive products for human reps. Held at 4 rather than 5 because no containment or resolution rate is published, so the size of the end-to-end claim cannot be checked.Evidence for this score | 25% | 4/5 |
| Evaluation and QAThe platform overview names LLM-based evaluation, simulation and regression testing before release, and a separate quality-management product reviews live conversations. Held at 4 because no evaluation methodology or benchmark result is published in detail.Evidence for this score | 20% | 4/5 |
| Enterprise readinessPresented as a single cloud platform layered onto an existing stack, with guardrails and real-time moderation described in general terms. No on-premise or private-cloud option, data-residency control, or named compliance framework appears in public material.Evidence for this score | 20% | 3/5 |
| Integration depthThe integrations page names around forty systems across telephony and CCaaS, CRM and case management, knowledge sources, and the data warehouse. This is the best-evidenced criterion on the entry: the list is specific enough to check against a buyer's own stack.Evidence for this score | 20% | 5/5 |
| Published evidenceIntegrations and named customer deployments are documented in public. Pricing is not published at all, and the compliance badges on the site do not name the frameworks they stand for — so two of the things a buyer checks first cannot be checked here.Evidence for this score | 15% | 3/5 |
| Weighted total | 7.7/10 |