Workflow design
We map the work that's eating your team's time, identify which parts are AI-ready, and design a workflow that compounds — not one that needs babysitting.
Bespoke AI workflows for the work that's eating your team's week.
Quote drafting, customer reply triage, content generation, document processing, internal research. The repetitive workloads worth automating — designed, built, integrated, documented, and watched.
SAMPLE
- Enquiry emailFrom the website formIN
- Agent drafts the quoteYour templates, your toneAI
- You reviewNothing sends without youYOU
- Quote sentLogged against the CRM recordDONE
AI COST THIS MONTH$14 / $50 cap
the demo → the workflow
AI hype is everywhere. AI in your operation is rare.
Closing that gap is the work. Most AI projects die in it. Ours don't, because we treat the workflow as the deliverable, not the agent.
- You've seen the demos.
- You know AI can write the email, draft the quote, summarise the meeting.
- The gap is the workflow that actually runs in your business — every day, on real data.
- Integrated with your tools, with the guardrails that stop it going off the rails.
Six things, every workflow build.
Designed, built, integrated, documented, and watched.
Custom agent build
Claude Projects, Custom GPTs, n8n flows, or bespoke API agents — whatever fits the workload. Loaded with your business context, tested against real inputs.
Tool integrations
Connected to where the work actually happens — your CRM, your inbox, your spreadsheet, your booking system, your Slack. AI in the workflow, not in another tab.
Hard rules and guardrails
What the agent can do, what it can't, and what it has to ask a human about. Privacy boundaries, data-handling rules, escalation paths — all written down and enforced.
Cost monitoring
Token usage tracked per workflow. Monthly cost cap configured. Alert if spend trends past your threshold. AI runs are operational expenses; they get treated like one.
Playbook documentation
Every workflow ships with the Playbook that runs it — for the human steps and the AI steps. So when the team grows, training takes ten minutes, not three weeks.
Four phases. Three to eight weeks end-to-end.
Diagrammed before any code, stress-tested against twenty real inputs.
- 01
Audit the workload
What's currently eating time? Which parts of it are repetitive enough to automate? Which parts need a human? We list it before we build it. - 02
Design the workflow
Diagram the flow before any code. Where AI adds value, where humans stay in the loop, what the inputs and outputs look like. Reviewed with you before build starts. - 03
Build and stress-test
Agent built, integrations wired, tested against twenty real-world inputs. Edge cases written down. Errors handled. Cost tracked from day one. - 04
Launch and tune
First two weeks live, we watch the runs. Tune the prompts, adjust the rules, document what changed. After that you have a workflow that runs without us.
Kathryn leads every engagement.
Specialist contractors come in for design, copy, and bespoke development as the scope requires — all working off the same brief, all answering to the same quality standard.
A more organised business isn't about doing more. It's about removing the things the owner should not still be carrying.
Your questions, answered.
01What's an AI workflow build?
A bespoke CH Digitals engagement to design, build, and deploy a specific AI-driven workflow inside your business — quote drafting, customer reply triage, content generation, internal research, document processing, whatever the high-leverage workload is. Includes the agent build, tool integrations, hard rules, cost monitoring, and the Playbooks that document the whole flow.
02What does it cost?
Project-priced from a written quote after the discovery call. Most builds sit in the $5K–$20K range depending on complexity, integration count, and how custom the prompts and guardrails need to be. Token costs are billed separately at cost — we don't mark them up.
03How long does it take to build?
Three to eight weeks end-to-end. The variation is mostly about integration complexity. A workflow that lives entirely inside Claude or ChatGPT ships in three weeks. A workflow wired into HubSpot, Shopify, Slack, and your accounting tool takes longer.
04Will the AI go off the rails?
Not without us seeing it. Every workflow ships with hard rules, escalation paths, and audit logs. If the agent does something unexpected, it's logged and you're alerted. We don't deploy autonomous agents that act without a human review point — agency without accountability is how AI projects fail.
05What's an example?
Common ones: a quote drafter that reads a customer enquiry email, pulls the right templates from your CRM, writes a draft quote in your tone, and stages it for human review. A content generator that takes a product update and writes the blog post, the social posts, the email, and the Playbook changelog — all from one input. A research agent that summarises competitor pricing weekly. Each is a different build.
06How is this different from the Setup Session or the Playbooks?
The Playbook library is the recipes. The AI Setup Session gets your AI tools properly configured and documented. The AI Workflow Build is the bespoke layer on top — we build a specific workflow custom to your business, not from a template. Many clients do the Setup Session first, then a Workflow Build when there's a specific high-leverage automation worth investing in.
If you've identified the workflow worth automating — let's build it.
Free thirty-minute discovery call. We look at the workflow, the inputs, the integrations needed, and whether a custom build is the right fit. If a Setup Session or a Playbook would solve it instead, we'll say so.
Book a discovery call