Cognigy AI Solutions Architect

Ellison Coast Limited; Greater London; Hybrid; GBP 70,000 - 110,000; Full time

Job summary

Ellison Coast Limited is recruiting a NiCE Cognigy AI Architect to design end-to-end Cognigy solutions for enterprise customers in London. The role covers architecture, security, integration patterns, and deciding on agentic AI versus deterministic flows.

Occasionally hands-on work, you may build, test and debug with the team. You’ll ensure solutions are secure, scalable, and maintainable in production beyond go-live.

Qualifications

• 2–3 years designing and delivering conversational or agentic AI.
• Hands-on Cognigy architecture experience across Flows, Voice Gateway, and LLM nodes.
• Understanding of structured tool results, grounding, guardrails and agentic reasoning.
• Strong conversation design skills and production-ready mindset.

Responsibilities

• Own Cognigy solution design across customer use cases.
• Set architecture, patterns, and security controls for deployments.
• Decide when to use agentic AI vs deterministic flows or NLU.
• Build reusable patterns for prompts, tool schemas and integrations.
• Review engineers' work, unblock them, and sign off solutions.
• Run workshops, demos, design reviews and risk reviews with customers.

Skills

Cognigy architecture

LLM prompting

Tool orchestration

Conversation design

REST integrations

Security controls

Multilingual capabilities

Code review (Code Nodes)

Tools

JavaScript

Playbooks

Insights

xApps

Job description

NiCE Cognigy AI Architect (Conversational and Agentic AI)

London, hybrid (1 to 2 days a week in the office) | Permanent | Salary open, depending on experience

I’m recruiting for a NiCE partner that’s been around for over 20 years and is growing steadily on the back of a good pipeline. They’re building out a dedicated CX team with a Cognigy practice at its centre, and they need the person who designs the solutions the engineers then build.

You’d own the design end to end for customers. That covers the architecture, the standards, how integrations and security are handled, the conversation design approach, and final sign-off. So when an engineer asks whether something should be agentic or a straight deterministic flow, you’re the one who decides. This is still a hands-on role occasionally, though. If a project needs it, you’ll get into the platform yourself and build, test and debug alongside the team.

The real measure of the job is getting solutions into production for voice and chat that are secure, can scale, and can be maintained and tested long after go-live. A working demo isn’t enough.

What you’d be doing

• Owning the Cognigy solution design across each customer’s use cases
• Setting the architecture and patterns: flow structure, agentic AI strategy, tool calling, integrations, security controls and guardrails, multilingual, voice experience, xApps, reporting and production readiness
• Deciding when to use agentic AI, deterministic flows, NLU, Knowledge AI, xApps or a handover to a human
• Building reusable patterns for LLM Prompt nodes, tool schemas, Resolve Tool Actions, context, API integrations, fallback, escalation and error recovery
• Guiding the Cognigy engineers by reviewing their work, unblocking them when they’re stuck, and signing solutions off
• Running workshops, demos, design reviews and risk reviews with customers and internal stakeholders

What you’ll need

• At least 2 to 3 years designing and delivering conversational, voice or agentic AI. Cognigy’s only been used at scale for a few years, so this is about depth, not length of service.
• Hands-on Cognigy architecture and delivery experience, covering Flows, Voice Gateway, tool calling, LLM Prompt nodes, Knowledge AI, xApps, Playbooks, Insights, environment management and reusable patterns
• A practical understanding of LLM prompting and tool orchestration, including structured tool results, grounding, hallucination prevention and guardrails, and knowing where deterministic logic or NLU should sit alongside agentic reasoning
• Sound conversation design: intent journeys, task flows, fallback and repair prompts, turn-taking, escalation points, and wording that works when it’s spoken rather than read
• JavaScript that’s good enough to review Code Nodes properly, covering validation, payload mapping, API response handling, context, logging and custom flow behaviour
• Experience designing secure REST integrations, with PII-safe handling, authentication, error handling, observability and a sensible production support model
• Voice agent architecture: STT and TTS selection, multilingual recognition, DTMF, barge-in, end pointing, transfers, hang-up and silence handling, latency, and speech tuning
• The ability to get to the bottom of messy problems using transcripts, context, tool arguments, Voice Gateway logs, API responses, Insights and Code Node logs
• Ideally, a background at NiCE, a NiCE partner, or a CXone contact centre where you designed the Cognigy estate

Also useful

Cognigy certification. Deeper voice UX work, such as persona and tone of voice, dialogue repair, multilingual copy, usability testing, containment optimisation, or migrating an IVR to a voice agent. Contact centre architecture more broadly. Agent Copilot and Cisco Finesse integration. SSML and advanced TTS tuning. SIP, WebRTC, Twilio or similar. CRM, case management or CCaaS integrations. Experience putting together a guardrail framework covering prompt leakage, scope control, profanity filtering, hallucination prevention and speech-safe output.

What’s on offer

25 days’ holiday plus bank holidays, rising with service. A bonus scheme, a company pension, private medical cover, and a share scheme once you’ve been there a while. Also enhanced parental leave, a paid volunteering day, and funding plus study leave for qualifications.

A note on job titles

You might know this role as Conversational AI Architect, Agentic AI Architect, AI Solutions Architect, Cognigy Solution Architect or NiCE Cognigy Architect. The titles vary, but if you’ve been designing Cognigy solutions that went into production, this is the role you’re looking for.