During design and training you log requested changes in a shared Google Sheet, with an example when that makes the change clearer. We implement them and check them internally before we send the agent back. That runs across three review cycles.
AI agents
AI agentsAn agent built for one job.
We purpose-build AI agents for a defined problem: qualifying leads, help-desk replies, blog drafts, WhatsApp bookings, or a job that is yours alone. The point is measurable work your client can see, with people left on the high-value tasks. You keep the client.
How you know the agent is earning its place
PwC’s AI agent survey reports that nearly two-thirds of companies adopting AI in their processes see higher productivity, and that over half report cost savings, faster decisions, or a better customer experience. That is published research, not a result we promise for your client.
- 01
More time for high-value work
Repetitive work such as reporting, research, or data entry is handled by the agent, so people can stay on strategy, relationships, and growth.
- 02
Someone answers around the clock
From customer questions to lead qualification, the agent is available 24/7, without a full-time support team for that first line.
- 03
Capacity without a new hire for every task
Another agent can be added as the need grows, instead of hiring and training a person for the same repetitive job.
- 04
Decisions with less spreadsheet time
Insights are produced from the work the agent is doing, so the team is not assembling the same report by hand.
- 05
The same process every time
The job runs the same way on each pass, which removes the small errors that cost credibility or money.
- 06
A result you can point at
The brief names the outcome: more qualified leads, faster support replies, steady blog output, bookings, or traffic. Those are the measures, not a guaranteed return.
- 07
Routine cost down, the offer still premium
Moving routine work off the team reduces that overhead, while the client still receives a defined, tech-forward service.
Agents we can build
Each agent is for one challenge. A job that is not on this list can still be scoped as a custom agent.
Lead qualification
For sales time lost on unqualified leads. The agent screens inbound leads from web forms, social, or live chat, asks qualifying questions from criteria you set, scores the lead, and routes the qualified ones to the right person in the CRM or another channel the team already uses, such as Google Sheets.
Help desk
For support teams stuck on repeat questions. The agent answers common queries from the knowledge base and, for anything harder, opens a ticket and routes it to the right person. A 2025 industry article reports that AI-powered tools can reduce resolution times by up to 50%, and that agents can offer 24/7 support. That figure is not a result we promise for a help desk we build.
Blog drafts
For a blogging schedule that is hard to keep. The agent can watch RSS feeds or YouTube channels you name, summarise new material, and draft a unique, SEO-oriented post. The draft can be pushed to a CMS such as WordPress. A person still owns what gets published.
WhatsApp
For leads and bookings missed because WhatsApp is not staffed around the clock. The agent connects to a WhatsApp Business account, answers FAQs, shares business hours, pre-qualifies leads, manages appointment bookings, and pushes lead data to a CRM.
Custom
For a workflow that is not one of the four above. We design and build an agent around that agency or client process and the goal named in the brief.
How it runs
From the brief to launch, in four steps
A briefing document comes first. From that we design the strategy and send a phased quote and payment plan. Three review cycles sit inside design and training.
Three review cycles, then a separate quote
New functionality or a new integration is written up with an updated cost and timeline. A significant rework gets a revised quote and project plan before we continue.
What we do, and what we need from you
From the brief to rollout we run the build. Projects stall most often when logins, processes, or review notes arrive late.
| Task | Your agency |  |
|---|---|---|
| Review and analyse the brief you submit, including through Basecamp. | ||
| Keep the timeline, tasks, and milestones, and tell you when a stage moves. | ||
| Design the agent’s role, workflows, and conversation logic against the business goals. | ||
| Build, train, and refine the agent for accuracy, reliability, and scale. | ||
| Test it across real-world scenarios, including performance and responsiveness. | ||
| Hold regular check-ins so progress is visible and adjustments can be made quickly. | ||
| Document any change to the original requirements with an updated timeline and cost. | ||
| Validate the agent against the agreed KPIs and benchmarks before deployment. | ||
| A completed brief: goals, challenges, and what you expect the agent to do. | ||
| Logins or integrations for the CRM, scheduling tools, and knowledge bases, so the agent can be trained. | ||
| The processes, FAQs, or frameworks the agent should follow. | ||
| Consolidated feedback inside the scheduled review cycles. | ||
| Someone on the regular calls, so a blocker does not wait a week. |
How the work is billed
Projects are billed in two or more instalments. The first is billed when the project starts. After that we charge monthly, as scoped at the start. A part of the service is a one-off and is invoiced that way. Recurring work is charged from the day the recurring project is opened until the first of the next month, or until you pause it, and then monthly while it runs.
If requirements shift, new functionality or integrations are documented with updated costs and timelines. A significant rework gets a revised quote and project plan before we continue. Gather the client information before you open the project, so there is no gap between opening it and starting the work.
How requests stay visible
Distance and missed detail are handled with the working method, not with hope. We use live collaboration for screen sharing, testing, and progress, written updates so the state of the work is plain, and a change log so a request is not dropped.
You can add every change to the shared Google Sheet from the first review. If something is only said on a call, write it into the sheet so it is in the cycle.
How to offer the first agent
Clients buy an outcome they can measure, such as faster replies, lower routine cost, or more conversions. They do not need a tour of models. Start with one low-risk, high-value job, such as a lead qualifier, a reporting assistant, or an FAQ bot, and use that as the example for the next client.
A pilot is easier to approve than a large programme. We do not publish performance-based pricing. The quote is the phased plan from discovery: instalments, then the monthly scope. A story you tell your client should be a result you actually have, not a sample claim.
Agencies often add these around an agent
From the blog
Notes that sit next to this work. All notes
AI Solutions
How to Build an AI Marketing Strategy for Your Clients
A step-by-step approach to building an AI marketing strategy for clients, from auditing workflows to choosing use cases and measuring impact.
3 min read
AI Solutions
Using AI and Automation to Improve Customer Engagement
How brands can use AI agents, chat and message automation to respond faster without losing the human touch customers expect.
2 min read
Frequently asked questions
Agencies that want to offer a purpose-built AI agent under their own name, for their own operations or for a client.
Ready to brief the first agent?
You keep the client. We design, train, and test the agent, and we check it against the KPIs before launch.