Technology and applied AI
The dev practice inside Zenith Strategy.
Zenith Labs designs and builds two things: complex B2B customer acquisition systems, and custom agentic AI applications for mid-market businesses. Both are built to be handed over and run, not demonstrated once and left alone.
Customer acquisition
Complex B2B customer acquisition systems
Long sales cycles, small buying committees, and a handful of enquiries a month that genuinely matter. We build the whole path from first visit to booked conversation: the site, the pages behind each campaign, lead magnets that qualify rather than simply collect, and the sequences that follow. Instrumented end to end, so you can see which parts are earning their place.
What we build
Websites
Built on a modern stack and structured so marketing can change copy without waiting on a developer.
Landing pages
Campaign pages built to a system, so the next one takes hours rather than a fortnight.
Complex lead magnets
Surveys, self-assessments, benchmark tools and application forms. Instruments that qualify a prospect while giving them something worth the exchange.
End-to-end nurturing sequences
Drip programmes triggered by what someone actually did, segmented by what they took and how they answered.
CRM architecture and integration
Contact model, tagging, segments, and the API work that connects your forms to your marketing platform.
Analytics and attribution
Event tracking and reporting that ties a form submission to a closed engagement rather than to a click.
Technical SEO and performance
Core Web Vitals, structured data and indexing. The work that decides whether any of the above gets found.
The stack
Application
Next.js
TypeScript
Tailwind CSS
SanityEdge and hosting
Vercel
Netlify
CloudflareSurveys and capture
CRM
HubSpot
SalesforceMeasurement
GA4
Tag Manager
Meta PixelLinkedIn InsightFlywheelApplied AI
Custom development of agentic AI applications
Most mid-market AI projects stall because they begin with the model rather than the work. We start with a process that costs real money, then build the narrowest thing that changes it. On your own infrastructure, with your data staying where your policies require it to stay.
What we build
Proprietary AI workflows
Agentic workflows built around a specific process in your business, rather than a general assistant bolted onto it.
Custom MCP servers
Business-specific Model Context Protocol servers that give a language model controlled, auditable access to your own systems.
Model tuning and ongoing optimisation
Evaluation harnesses, prompt and model iteration, and monitoring that tells you when output quality drifts.
Retrieval over your own material
Indexing and retrieval across contracts, reports and operational records, so answers cite something real.
Human-in-the-loop review
Approval steps, audit trails and escalation paths, so an agent can act without acting unsupervised.
Line-of-business integration
ERP, CRM and finance systems connected through proper APIs rather than screen scraping.
Governance and cost control
Access control, logging, token budgets, and a clear view of what each workflow costs to run.
The stack
Application
.NET
PythonOrchestration and agents
Microsoft FoundrySemantic Kernel
MCPModels and inference
Anthropic
OpenAIxAI
Perplexity
Hugging FaceData and retrieval
Microsoft Fabric
Azure AI Search
pgvectorPlatform and operations
Azure
GitHubAzure MonitorWorking with us
The increasing convergence of strategy and tech
Zenith Labs was set up because the recommendations coming out of our strategy work increasingly needed someone to build them, and because the businesses we work with would rather not brief an agency from scratch. We also take Labs engagements on their own terms, for organisations that already know what they need built.