Look closer. Try it yourself.
What happens when you put a claim to the test? This is the place for research, comparisons and observations from real work.
Each investigation begins with a question. A post shows how the result was reached: the sources considered, tools tried and limits of the observation.
Keep reading
More questions. Practical answers.
Website
AI plans the shot. And the route behind it
How our developed drone-planning workflow connects light, camera actions and flight paths, with a documented test and human supervision.
AI portraits you can recognise. Checked with a system
How our developed portrait workflow selects references, measures facial similarity and filters weak results. Explained with three fictional characters.
AI sends staff out. And stays in touch
AI sends staff out with clear assignments, communicates with them and explains what to check or capture on site.
Knowledge that is still there next time
Our developed AI memory connects search, sources and project relationships, turning earlier work into usable context.
Project Bantayan: Built with AI. Maintained by AI
Our community project shows the breadth of our work: AI maintenance, custom software, a ferry view and data integration.
Reply to reviews. Keep your voice
How we connect AI replies with a clear response policy and approval for criticism: inside the developed workflow.
Turning research into a decision
Our research atlas connects sources, risks, judgements and implementation. Drone planning shows how research leads to action.
What an AI visibility check actually measures
Research note, 19 November 2026: what mentions, citations, visits and enquiries tell you, with the limits documented by Google and Bing.
Customer service
An AI agent that gets tasks done
Inside our developed customer-service agent: retrieving knowledge, using tools and involving people where needed.
Discussed on the phone. Ready in an email
How our AI phone workflow follows up a discussed offer by email when the caller explicitly requests it.
Following up missing documents without blind reminders
Our AI-assisted reminder development checks outstanding requests and documents already received before taking the next step.
FRED and HubSpot: the case behind the message
How we connected HubSpot to our AI agent: using contacts, tickets and communication in context instead of piecing them together repeatedly.
FRED and Stripe: invoices from the right source
Our Stripe integration connects an AI agent to current invoice data, turning a payment question into a traceable answer.
Helping FRED keep its knowledge current
How we built team knowledge maintenance with AI: review existing material, justify changes and approve them deliberately.
One case. Several sources. An evidence-based overview
How we developed internal AI research for staff: connecting sources, highlighting contradictions and preserving unanswered questions.
Reviewing fraud signals with local AI
How we used AI to build a fraud-review system: a local model, connected accounts and understandable signals for human review.
Turning an AI phone call into an actionable case
How we developed the handover from a FRED call to the team: request, context and a confirmed case-creation result.
Turning an email into a reviewable case
How our AI-assisted email workflow connects information, a draft response and the human actions a case needs.
Updating a payment method from the conversation
How FRED guides a verified customer to the relevant Stripe self-service flow without collecting card details in chat.
When HubSpot changes, the next task can begin
How our event integration receives HubSpot changes, retrieves context and prepares the next stage of AI-assisted work.
Something to read today
Start with understanding.
Find foundations and practical questions in Knowledge. Articles stay separate from future experiments and accounts of our work.
What does this mean for your task?
Reading helps you put things in context. For your own workflow, tell us about the task or goal.
