Agentic voice, email and chat on the contact centre you already run, whether that is Genesys, Avaya, Amazon Connect or something else. We take pilots through the ten production gates that stand between a working demo and live traffic.
15 questions. 3 minutes. Free. Your recommendations immediately.
Delivering for a client rather than buying for your own contact centre? We work under your brand.
Trusted by teams at
AWS Advanced Partner | Amazon Connect Delivery Partner | Top 3 Agentic AI Partner in EMEA
Four agents running in production, one per industry. Voice is the quickest way to judge one, so start there. No signup, no sales pitch.
Chat and email run on the same account data and the same rules. Try all three per industry
Calling from outside the UK? Nova also answers on +1 469 868 9323 in the US and +61 1800 616 841 in Australia.
Each channel gets its own agent, because voice, chat and email fail in different ways. They share your account data, your integrations and the same ten gates. Nobody else in this market publishes what a build costs. We do.
Bronze
£25,000
One channel, into production. 4 weeks.
Most operations need this one
Silver
£55,000
A channel, fully covered. 6 to 8 weeks.
Gold
£120,000
Agentic CX across channels. 8 to 12 weeks.
Every package runs on the contact centre you already have. You do not migrate: same numbers, same routing, same agents, with the agent layer on top.
What each package includes, and how you payLive systems, named where the client has cleared it, with the scope and the numbers behind each one.
Specialist agents for strategy, contact, negotiation and real time compliance monitoring, on Strands Agents and Bedrock across voice, SMS and email. Multi tenant, with FCA Consumer Duty controls in the conversation flow. AgentApex AI Ltd is our sister company, named with our own permission.
Europe's largest credit management company. An autonomous email agent reads every inbound message, classifies it, creates a case and routes it to one of 13 team queues. 2,400 a month, FCA compliant, full audit trail. On Strands Agents, Bedrock and AgentCore with Amazon Connect Cases. Live since 2026.
Ten gates every deployment has to pass before it carries live traffic, across three areas. Most contact centre AI stalls because none of them were in place, not because the model was wrong.
These are the same ten questions the assessment scores you against. The full definitions, with the exclusions written in, are published on how we measure, and the limits we have not solved are on what breaks.
Production readiness 4 gates
Fallback when the model is unsure, p95 latency under load, an evaluation suite on every release, and a handover that carries full context.
Measurement discipline 3 gates
A cost per contact baseline before go live, containment with hang-ups and 48 hour repeat contacts taken out, and resolution counted only against a matching write in your system of record.
Compliance posture 3 gates
Identity verified before an account is discussed, vulnerable customer guardrails tested inside a live call, any decision replayable months later, and a security review passed with findings closed.
The six reasons this conversation usually stops, answered before you have to ask. Every answer is something you can check on this site.
Then keep it. The agent layer runs on AWS alongside your contact centre, not instead of it: same numbers, same routing, same agents. Platform migration is explicitly excluded from every package, because you do not need it to get agents.
It hands to a human when the model is unavailable or unsure, and an evaluation suite runs on every release before anything reaches a customer. Both are production gates, not options. Email can start in draft mode, where your team still sends.
Vulnerable customer guardrails are documented and tested inside a live call, any individual decision is replayable and auditable months later, and a security review is passed with the findings closed. The definitions are published, with the exclusions written in.
You are not buying a second one. The agent layer sits on top of what you run. Platform licence costs for Amazon Connect, Bedrock and telephony are billed by AWS to your account, not through us.
Four weeks is Bronze: one channel, one use case end to end, with a cost per contact baseline captured before go live. It is not three channels, ten use cases or a migration. Those are eight to twelve weeks, and priced separately.
Everything is in your repository. You get runbooks, dashboards, a handover walkthrough and an evaluation suite your own team runs. The engineer you interviewed before the work started is the one who hands it over.
You sold it. The client is waiting. The people who can build it are not on your bench.
Engineers embedded in your delivery team, on your client relationship, reporting through your leads. The client sees your team.
Not during the engagement and not after it. We will sign your non-solicitation terms rather than ask you to accept ours.
Published day rates, four engagement shapes, and people who have put agentic systems into production in regulated environments.
"Most AI projects fail because they never leave the lab. We only build what we can put into production. Every agent we deploy handles real customers, real conversations, real money."
You interview every engineer before they start, and the one you meet is the one who shows up. How we staff an engagement
Fifteen questions, scored against the ten gates above.
Find out what is blocking you15 questions. 3 minutes. Free. Your recommendations immediately.
Already know what you need? Book a discovery call or email hello@rel8.cx