AI agent packages

AI agent packages that survive contact with production

Start with one use case and real evidence. Expand only when the numbers justify it.

Review first Go or no-go we say no when the numbers say no

AI Agent Pilot

One agent on your real data, with honest numbers on whether to scale it.

What's included

  • Use case selection workshop scored on value and feasibility
  • One working agent built against your real data
  • Evaluation suite so quality is measured, not assumed
  • Guardrails and human escalation paths
  • Live testing with a small internal group
  • Cost per interaction modelled against the current manual cost
  • Go or no-go recommendation, including no-go if that is the honest answer

Not included: Production rollout to customers · Multiple use cases · Model fine-tuning

Yours to keepWorking agentEvaluation resultsCost modelWritten recommendation
Full build ~5x cheaper per conversation than a human-handled ticket, on published benchmarks

Support Deflection Agent

Routine questions answered around the clock, with clean handover when a human is needed.

What's included

  • Knowledge source ingestion from docs, tickets, and knowledge base
  • Retrieval grounding so answers cite real sources instead of improvising
  • Agent built on your platform of choice: Agentforce, Breeze, or custom
  • Channel deployment across web chat, email, or messaging
  • Escalation with full conversation context to human agents
  • Evaluation suite and hallucination testing before launch
  • Deflection, containment, and CSAT dashboards
  • Tuning across the first month of live traffic

Not included: Telephony or voice channel build · Platform licence costs · Content writing for missing knowledge

Yours to keepProduction agentEvaluation suiteDashboardsTuning runbook
Custom build 92% autonomous resolution on a multi-agent build Salesforce called impossible

Legacy System Retirement Agent

An agent working against the same data, so the old interface can finally be switched off.

What's included

  • Discovery of what the legacy system actually does versus what people believe it does
  • Data model and access mapping
  • Custom agent built on the LLM that fits the workload
  • Tool and action layer against the underlying data
  • Full evaluation, observability, and audit trail
  • Phased cutover with parallel running
  • Team enablement and change management

Not included: Data centre or infrastructure migration · Retirement of systems with hard regulatory retention requirements without a compliance review

Yours to keepProduction agentObservability stackCutover planAudit documentation
Ongoing support Measured against a maintained test set, not assumed

Managed AI Agents

Continuous evaluation, so quality holds as products, policies, and data move.

What's included

  • Continuous evaluation against a maintained test set
  • Prompt, topic, and action tuning
  • Knowledge source refresh
  • Quality, cost, and containment monitoring
  • Incident response when an agent misbehaves
  • Monthly performance review

Not included: Model provider costs · Net-new agent builds, scoped separately

Yours to keepMonthly quality reportContinuous tuning

In every package

The floor is the same, whatever you pick

A written scope

Everything in and out, agreed before we start.

A fixed price

Quoted against that scope. Changes are priced separately, and only if you want them.

Senior architects

People who ship these builds every week, not juniors learning on your project.

Thirty day hypercare

We stay close after go live, then hand over documentation you can actually use.

Packages on other platforms

Same fixed-scope model, different logo

Back to all packages
FAQ

Questions about AI Agents packages

Fixed scope, fixed price

Not sure which AI Agents package fits?

Tell us what is not working on a free thirty minute call. We will tell you which package fits, or that none of them do and what you need instead.

Ask me anything