AI workflows
AI workflows to run your website from chat
An AI agent your team asks in Slack or Claude drafts pages, content and reports on your design system. Nothing goes live until you approve it.
Built by the team behind sites for
How it works
Your team can Ask in chat, the agent Drafts it with your components, and you Approve the preview before it goes live.
Can you add a pricing page for the new plan?
- Reading the brief and brand rules
- Pulling components from the design system
- Drafting the page in Astro
- Checking contrast, spacing and links
- Optimising images and speed
Approved by you · Live
Our setup process
From busy backlog to workflows your team uses.
We start with the requests that eat your week, prepare the site and design system the agent will use, then connect, test and hand over.
Week 1
Workflow map
We list your team’s weekly requests, pick the ones worth automating, and agree each one’s output, sources and approver.
Typical outputs:
- 01Request inventory
- 02Workflow shortlist
- 03Owners and approvers
- 04Success measures
Week 1-2
Foundations
An agent is only as good as what it builds with. We tidy the design system, sections and CMS so drafts come out on brand.
Typical outputs:
- 01Component and section rules
- 02CMS content model
- 03Brand and tone guide
- 04Example requests
Week 2
Connections
We connect chat, the site or repo, the CMS and your data, with separate rights to read, draft, publish and approve.
Typical outputs:
- 01Slack or Claude setup
- 02CMS and repo access
- 03Analytics sources
- 04Permission map and logs
Week 2-3
Pilot
We run the workflows on real requests, check the drafts against your standards, and fix what the agent gets wrong.
Typical outputs:
- 01Tested workflows
- 02Preview and approval flow
- 03Known limits list
- 04Before and after timings
Week 3-4
Handover
We train the people who ask, review and approve, document every workflow, and agree who looks after it.
Typical outputs:
- 01Team training
- 02Workflow playbooks
- 03Failure and recovery plan
- 04Next workflows backlog
What we deliver
Useful workflows, not another AI demo.
A small set of workflows your team will actually use every week, built on your design system and your rules.
A clear workflow map
Each workflow’s owner, sources, output and approval step, written down before the build.
An agent in your chat
Requests start in Slack or Claude, and drafts and previews come back to the same thread.
On-brand by design
Pages built from your approved sections and tokens, so drafts look like your site.
Scoped connections
Access limited to the agreed work, with separate rights to draft, publish and approve.
Preview before publish
Every change comes with a preview link. Your approval publishes it, never the agent.
A team that can run it
Training, playbooks and a known-limits list, so the workflows keep working after we step back.
Why DesignMe
Designers set the guardrails. The agent does the busywork.
Most AI setups fail on taste and structure, not on the model. We design the system the agent works inside.
- We run our own site this way
- Our own website is built in Astro and edited with Claude: we ask for pages and fixes, review the change, and approve it. We set up what we use ourselves.
- The design system comes first
- We build or tidy the components, tokens and content model before connecting an agent, which is what keeps its output on brand.
- Any stack, honest about limits
- Code-based sites (Astro, Next.js) give agents the most room. Framer and Webflow now connect to external agents too. We tell you what works on yours.
- Your approval, every time
- Drafts, previews and logs by default. We design the permissions so the agent can help without being able to surprise you.
- Measured on real requests
- We test with your actual backlog and agree what to measure, then keep the workflows that save time and drop the ones that don’t.
Tools and workflow
Works with the tools you already use.
Slack or Claude as the front door, your site and CMS as the workspace, and the data sources you trust for reports.
Chat and agents
- Slack
- Claude
- ChatGPT
- Notion
Websites and CMS
- Astro
- Next.js
- Framer
- Webflow
- Sanity
Data and reports
- Posthog
- Google Analytics
- Search Console
- Hotjar
Design and delivery
- Figma
- GitHub
- Linear
- Loom
Testimonials
Words from the teams we’ve worked with
“Working with DesignMe feels like having a design team embedded inside Quotient. Their design sense, responsiveness, and ability to iterate quickly make them an ideal partner as we grow.”
- Raised by Quotient
- $5M
- Embedded growth partnership
- 18+ months
Max DavishCo-Founder and CTO“They were very willing to attempt and deliver on anything we threw at them, which is what you want from a subscription based service where you're paying for continued deliverables.”
Mike FoggCo-founder @Heroes (prev. Blockview)
Faqs
Your questions about AI workflows
Still having more questions?
Tell us which requests eat your team’s week. We’ll tell you which ones are worth handing to an agent, and which aren’t.
What can we ask the agent to do?
Typical workflows: draft or update CMS content, build a new page or section from your existing components, make small site changes, and pull reports from connected analytics. We agree a short list with you, so the scope is clear from day one.
Can it publish to the live website?
Only through the approval step we set up. The agent prepares the change and a preview; someone on your team approves it, and that approval publishes it. Who can approve what is part of the permission map.
Is this an AI website builder?
No. It helps your team run an existing website through chat, using your design system, code and CMS. If you need those foundations, our website design and development services build them first.
Which tools and platforms do you support?
Slack or Claude for requests; Astro, Next.js, Framer and Webflow sites; headless CMSs such as Sanity; and analytics such as PostHog or Google Analytics. Code-based sites give the agent the most room. We confirm what fits your stack during the workflow map.
Can you add this to a site another team built?
Often, yes, after a short review. The agent needs a clear component structure and CMS model to produce good work. If those are missing, we fix them first, and that is usually where the time savings come from anyway.
Where does it run, and what can it access?
On your accounts and infrastructure, with access limited to the agreed workflows. Reading, drafting, publishing and approving are separate permissions, and every action is logged.
What happens when it gets something wrong?
It happens, which is why nothing goes live without review. We agree how errors are reported, when a request is retried, and who steps in. The pilot is there to find the weak spots before your team depends on it.
Does it replace our designers or developers?
No. It takes the repetitive requests off their plate. New design direction, strategy and complex builds still need people; the agent makes the everyday changes faster.
Do we need an ongoing retainer?
No. The workflows are yours and run on your accounts. Many teams keep us on to add new workflows or pair it with a design subscription, but that is optional and scoped separately.
Do you guarantee time savings?
We don’t promise a number. We measure real requests before and after the pilot, agree what success looks like, and keep the workflows that earn their place.
You've reached the end, so now…
Tell us what you’re building.
We’ll help you map the right scope, team, timeline, and next steps.
Book a call