AI tools aren't the deliverable. Working systems are.
Custom-scoped autonomous agents injected into your operations. Intake, ops orchestration, sales follow-up, vertical-specific workflows, designed against your business, built to a measurable outcome, guaranteed for 90 days.
An autonomous system, not a tool your team has to use.
Most AI offerings are tools. A chatbot you have to configure. A model you have to prompt. A workflow your team has to remember to trigger. Agentic injection is different. We embed agents into your operations that run continuously, handle the work end-to-end, escalate only the decisions that need human judgment, and report on what they did. The system is the deliverable, not the tooling.
Intake agents
Handle initial customer enquiries across WhatsApp, email, web chat. Qualify, book, and hand off to a human only when judgment is needed. For multi-branch operations, route by branch, location, and capacity automatically.
Ops orchestration agents
Monitor job pipelines, flag bottlenecks, reschedule automatically when something slips, and report daily to ownership. The kind of work an operations manager does, running 24 hours a day without burning out.
Sales follow-up agents
Run multi-touch sequences for warm leads who haven't converted. Sentiment analysis to detect interest signals. Human handoff at the right moment. Recovers revenue most teams have already written off.
Vertical-specific autonomous workflows
Built against the specific operational pattern of your business. Funeral parlour intake plus FICA collection plus family notification plus policy reminder. Auto repair intake plus insurance panel routing plus parts ordering. Bespoke to what you actually do.
Back-office decision agents
Categorise invoices, route approvals, flag anomalies in spend, generate the routine reports a junior team member would otherwise produce. The work that scales linearly with revenue and shouldn't.
Custom-scoped systems
If your operational reality doesn't match any of the above patterns, we scope from scratch. Discovery is real engineering analysis, not a templated questionnaire. The system is designed around your actual process.
Hit the agreed outcome in 90 days, or we work for free until we do.
Discovery quantifies the outcome. We agree on a measurable result against the process being automated, recovered hours per week at a senior level, autonomous handling of N percent of qualifying enquiries, response time below a target threshold across the full pipeline, or a process-level metric that reflects the actual business problem being solved.
That number is locked in writing before you pay a cent. From the day the system goes live, you have 90 days. If the agreed outcome isn't hit by day 90, we keep working at no further cost until it is. Real engineering gets real time to prove out, with real accountability when it doesn't.
Why we can promise this: we deliberately cap intake at one to two new AI Systems engagements per quarter. These are real builds with real engineering effort. Overbooking would mean missing outcomes for everyone. The cap is what makes the 90-day promise honest.
Project fee plus monthly retainer.
The build is real engineering work and is priced as such. The retainer covers ongoing model and tooling costs, eval maintenance, and new-workflow additions within scope. Both numbers are scoped in discovery against the specific work; the ranges below are what an SA mid-market SME typically commits in Phase 1.
Project fee
Discovery, architecture, build, integration, eval, and tuning. Paid 50% on signature, 50% on go-live. Scoped against the specific work in discovery.
Monthly retainer
Model and tooling cost passthrough, ongoing eval maintenance, weekly tuning, and new-workflow additions within scope. Paid monthly. Cancel with 30 days notice.
Total first-year value (typical)
Project fee plus 12 months retainer. The actual number depends on scope; this is the band an SA mid-market AI Systems engagement typically lands in.
If your operation is smaller and you'd rather have a productised retainer with predictable monthly costs, look at AI Automation. If you don't have a website yet, that's where we'd start.
How an engagement runs
Discovery (1 week)
Ty sits with your team. Documents the process to be automated. Quantifies the outcome we'll be measured against. Identifies integration points, data flows, edge cases, and human-in-loop requirements. End of week you have a written discovery report and a ballpark scope.
Architecture and scoping (1 week)
We design the agent system, write the technical spec, and quote a fixed scope and fixed price. You get the spec, the price, the agreed outcome metric, the 90-day window, and the work-for-free clause. 50% deposit on signature.
Build (2 to 4 weeks)
Build, integration, and testing happen against your real conversation history or process data. Weekly progress demos. You see what's being built and provide feedback as it takes shape, not at the end.
Eval and guardrails (1 week)
We test failure modes systematically. Build human-in-loop fallbacks for every edge case identified. Verify accuracy thresholds against agreed metrics. Document where the AI is confident and where it escalates. The system isn't ready until it fails gracefully.
Tuning, handover, and 90-day window
System goes live with monitoring. 50% balance due. Your team is trained on the override and feedback mechanisms. From this day we're measured against the agreed outcome. We tune weekly. At day 90 we hit the metric or we keep working at no further cost.
Questions executives ask before booking discovery
How is AI Systems different from AI Automation?
AI Automation is a productised retainer ladder that handles inbound, follow-up, and CRM logging using off-the-shelf adapters per vertical. It's right for SMEs receiving up to 200 leads per month who want predictable costs and quick setup. AI Systems is a bespoke engineering build, custom-scoped agents embedded into your operations, designed against your specific business processes. It's right when the off-the-shelf retainer doesn't fit because the work that needs automating is unique to how your business actually runs.
What is "agentic injection" in plain English?
An autonomous AI system that does the work, end-to-end, instead of giving your team a tool that helps them do the work. The difference matters. A chatbot is a tool. An agent that handles incoming customer enquiries across WhatsApp, email, and chat, qualifies them, books them in, escalates the ones that need human judgment, and reports daily, is a system. The second is what gets built here.
What does an engagement actually look like?
Discovery (1 week): Ty sits with your team, documents the process to be automated, and quantifies the outcome to be measured against. Architecture and scoping (1 week): the agent system is designed, the spec is written, and sign-off captured. Build (2 to 4 weeks): the system gets built, integrated, and tested against real conversation history or process data. Eval and guardrails (1 week): failure modes are tested, human-in-loop fallbacks built for every edge case, accuracy thresholds verified. Tuning and handover (1 week): the system goes live with monitoring, your team is trained on the override and feedback mechanisms, and the 90-day outcome window begins.
How does the outcome guarantee work?
Different from the AI Automation guarantee in two ways: the outcome we agree to is bigger (typically a process-level metric, not a single workflow metric), and the window is longer (90 days, not 60). The mechanic is the same, we quantify the outcome in discovery, lock it in writing, and if it isn't hit by day 90 we work for free until it is. Real engineering gets real time to prove out.
Why does pricing have a range?
Because the work has a range. A single-process agent (e.g., intake automation for a multi-branch operation) is at the lower end. A full ops orchestration system spanning four processes with custom integrations into legacy software is at the upper end. The discovery call is where we scope and price specifically. The R30K to R150K project range covers what an SA mid-market SME typically commits upfront in Phase 1; if your scope is materially larger we'll quote separately and won't pretend it fits the productised range.
Who owns what at the end?
You own everything: the agent code, the prompt library, the integration credentials, the eval suite, the monitoring dashboards. The retainer covers ongoing tuning, model and tooling cost passthrough, eval maintenance, and new-workflow additions within scope. If you cancel, you keep the system. We hand over operationally and step out. We've designed it so you're never trapped.
How is this priced beyond Phase 1?
We're transparent about this: the flat project-fee plus retainer model fits Phase 1 because it's predictable for both sides while we learn what an agentic system is worth in the SA market. We've committed to revisiting the pricing metric after 5 to 10 deals, the likely evolution is per-agent-deployed or per-process-automated, which captures upside as systems scale. If you're an early Phase 1 client, you're locked in on the simpler model regardless of how it evolves.
Ready to see what an agentic system would look like in your operation?
A free 60-minute discovery call. Ty walks through your operational reality, identifies the highest-value processes to automate, and gives you a ballpark scope before any engagement. No deck, no pitch, just analysis.
Book a discovery call