How We Build · Safe

Every claim on the site traces to something you said.

A BeargrassAI build is not a template. It is a short, honest process built around one rule: every claim on the site traces to something you said. Here is how that actually works.

I · The process

Five steps, in order

  1. Intake conversation. Thirty to sixty minutes, in person when you are in the valley, on a call otherwise. We talk about your business, your customers, and what a customer should be able to find or do on your site.
  2. Claims extraction. Mark drafts a claims file from the conversation and your existing materials. Hours, services, prices, policies, specialties. You review every claim before it lands.
  3. Build. Semantic HTML, modest CSS, Cloudflare hosting. Your Google Business Profile is tuned in parallel. If you ordered an Agent, the Agent is wired to the claims file as the only source of truth.
  4. Validate. Lighthouse performance, accessibility, structured-data validation, and a Bloom Test run on the new site. You read the site. Mark reads the site. Everything changes until both of us are satisfied.
  5. Ship and maintain. DNS flip, one short handoff, then quiet maintenance. Infrastructure kept current, monitoring in place, light content updates included. Big changes are a conversation.
II · Why it works

Depth over volume

BeargrassAI keeps a small number of clients. You can see a few of them in a single sentence each on the About page. That depth is the whole model. A business we build for is one we want to understand well enough to speak for, accurately, in public. Mark has turned away work that would have thinned the claims file. The work is simple. The attention behind it is not.

III · The claims file

One source of truth, two outputs

A list of sworn-testimony statements about your business that you approved. Hours, services, prices, policies, specialties, coverage area, payment methods, allergen practices for a kitchen, emergency coverage for a clinic. Anything a customer might ask and anything the site or the Agent might say. Every claim is categorized (foundation, service, pain, trust, audience, location, differentiation) and every claim is confirmed with you before it leaves your desk.

The claims file is the source of truth for two separate outputs at once. The site is built from it. The Agent reads from it. When a claim changes (new hours, new service, new price), we update the file once and both outputs reflect the change. No drift between what the site says and what the Agent says.

Start the intake conversation