AI operating assistant case study
Building Briefable: a connected AI operating assistant for founders.
How Changeable designed and built Briefable (formerly Stillwork), a modular AI platform that connects email, calendar, files, contacts and pipeline into one governed daily briefing, with human review kept visible before anything leaves the platform.
Project overview
A connected assistant, not another dashboard to check.
Founders and solopreneurs already use Gmail, a calendar, a CRM and a shared drive. What they lack is something that reads across all of it and tells them what actually needs attention today. Changeable built Briefable to be that layer: it connects to the tools a business already runs on rather than replacing them, and uses AI to organise, prioritise and prepare, while a human stays firmly in control of anything that leaves the platform.
The platform is modular by design. A small business gets a core set of daily-operating modules; specific add-ons like content production or performance analytics are configured only for the businesses that need them, rather than every feature being switched on for every user from day one.
Human review, always visible
Briefable drafts and prepares. Nothing is sent externally without a person approving it first.
Connects, doesn’t replace
Google Workspace, Microsoft 365 and Notion stay the source of truth. Briefable reads and organises across them.
Genuinely modular
Core modules ship with every deployment. Add-ons like content production are configured per client, not bolted on.
Multi-tenant from day one
Every table and every AI response is scoped to the individual user, so adding a new client never means forking the platform.
The problem
Founders drown in the space between their tools, not inside any one of them.
A solo operator or small team already has email, a calendar, task lists, a CRM and a shared drive. Each tool does its job well in isolation. What none of them do is tell you, at the start of the day, what actually matters across all of them together: which email needs a reply before a meeting, which deal has gone quiet, which task should genuinely come first.
Existing single-purpose tools solve one slice of this. Full executive-assistant AI agents solve the whole problem but at a price point built for larger teams, not a sole operator juggling everything themselves.
What most founders experience
- Starting the day by manually checking five different tools before knowing what to prioritise
- A CRM deal that quietly goes cold because nobody was prompted to follow up
- Meeting prep pulled together from memory rather than the actual history with that contact
- Single-purpose AI tools that each solve one problem, none of them talking to each other
- Full AI executive-assistant products priced for a team, not a solo operator
What Briefable does today
Seven core modules, two configured additions
Every deployment ships with the same core operating layer. Content production and analytics are available as configured additions for businesses that need them.
Today
A unified daily view combining calendar, tasks, email and an AI-generated brief.
Tasks
Voice and text capture, AI prioritisation and automatic subtask breakdown for complex work.
Inbox
AI-prioritised email with draft reply support and automatic action-item extraction.
Calendar
Multi-calendar awareness with meeting attendee context and focus-time protection.
Briefable
A full AI conversation with retrieval across the connected knowledge base.
Pipeline
A CRM surface with deal health scoring and stale-deal alerts built into the daily brief.
Contacts
Automatic contact capture from email and calendar, with follow-up reminders and health tracking.
Publish & Analytics
Configured additions for content production and performance reporting, delivered as a consultancy service.
How it was built
A modular build, proven module by module
Rather than build every feature at once, Briefable was developed one working module at a time, with the platform’s architecture holding steady while capability was layered on.
Multi-tenant, module-based, one AI layer
Before any feature was built, the foundational architecture was decided: Google Workspace as the backbone rather than a replacement for it, a single consistent AI layer for all intelligence rather than mixing models feature by feature, and a strict multi-tenant data model where every table is scoped to the individual user from day one. These decisions meant that adding a new client was always a configuration change, not a rebuild.
Today, Tasks, Inbox, Calendar, Knowledge and Pipeline
The core modules were built and connected in sequence: authentication, the unified Today dashboard, task capture, then Gmail, Calendar, a Drive-backed knowledge base and a CRM pipeline surface. Each module was tested end to end with real data before the next was layered on top, so the platform always remained a working product rather than a set of disconnected features.
Make the modules aware of each other, not just individually useful
Once the individual modules worked, the real differentiation came from connecting them: a morning brief that pulls from calendar, email, tasks and pipeline together; meeting preparation that gathers context from contacts, email history, internal knowledge and active deals in one pass; stale-deal alerts and capacity warnings surfaced automatically rather than requiring the founder to notice a pattern themselves. This is the layer that turns five separate tools into one operating assistant.
Content production and analytics as deliberate additions
Content production and performance analytics were built as configured additions rather than features every user gets by default. Content production connects to a business’s own writing voice, brand guidelines and publishing platforms; analytics pulls real search performance and content data for businesses that want it. Both are delivered as a guided setup, not a self-serve toggle, matching how deeply they need to reflect an individual business.
From an internal build to a product other people could trust
Once the platform was proven internally, it was rebranded from its working name to Briefable, with a full visual and naming sweep across the interface, domain and connected services. Before any external tester was invited in, a dedicated security review was run: every backend function was checked for proper authentication, unused endpoints were removed rather than left live, and access permissions were verified rather than assumed. Privacy and terms documentation was drafted directly from the platform’s actual data flows, and the sign-in experience was moved to a dedicated application domain separate from the marketing website.
The AI layer
Prepared, not just summarised.
Briefable’s AI does more than list what happened. Before a meeting, it gathers context from contacts, recent email threads, relevant internal knowledge and active pipeline deals, then produces a structured brief covering who the person is, recent activity, outstanding items and suggested talking points, rather than leaving the founder to piece that together from memory.
The same discipline applies to task management: complex tasks are automatically broken into a handful of concrete subtasks, and the day’s task list can be reordered by AI based on real calendar and pipeline context, not just due dates. Every one of these capabilities stops short of taking action on its own, drafts and suggestions always wait for a human to approve them.
Security and readiness
A dedicated review before a single external tester was invited.
Before Briefable moved from an internal tool to something other people could sign up for, a full security review was run against the codebase, not just the interface, informed by the same discipline Changeable applies in client AI governance work.
Results
What the finished platform delivers
The build produced a genuinely connected operating layer, not a collection of separate integrations.
One brief, five sources
The daily brief draws from calendar, email, tasks, notes and pipeline in a single pass rather than five separate checks.
Meeting prep in seconds
A structured brief on any upcoming meeting is generated from real connected context, not memory.
Nothing falls through silently
Stale pipeline deals and overdue follow-ups are surfaced automatically rather than depending on someone noticing.
Modular by design
New clients are a configuration change, not a rebuild, thanks to the multi-tenant architecture built in from day one.
Security reviewed before launch
A dedicated audit found and closed real authentication gaps before any external tester touched the platform.
Rebranded without disruption
The full platform, data and working experience carried through the rename from Stillwork to Briefable intact.
Questions
Questions about how Briefable was built?
Common questions about the approach behind Briefable’s connected AI operating assistant.
Why connect to existing tools instead of building a new inbox or calendar?
Founders already trust and rely on Gmail, their calendar and their CRM. Replacing them creates migration friction and risk. Briefable connects to what a business already runs on and adds an intelligence layer across it, rather than asking anyone to change how they work.
Does Briefable send emails or make changes on its own?
No. Briefable drafts replies, prepares briefs and suggests actions, but a human always approves anything before it goes external. This human-in-the-loop principle was a non-negotiable design rule from the first architecture decision.
Why was the platform built as modules rather than one fixed app?
Different businesses need different capability. A modular architecture means core operating features ship to every user, while content production or analytics can be configured only for the businesses that actually need them, without forking the codebase for each client.
Why was a full security audit run before external users were invited?
Multi-tenant AI platforms handling email and calendar access carry real risk if authentication is not correctly enforced everywhere. Running a dedicated audit, rather than assuming settings were correct, found and closed genuine authentication gaps before anyone outside the build had access.
Why was the product renamed from Stillwork to Briefable?
The name needed to reflect what the product actually does for the person using it: arrive at the start of the day already briefed. The rebrand was executed as a full sweep across the interface, domain and connected services once the platform was stable and proven.
What happens if the AI gets a priority or a brief wrong?
Every AI output in Briefable is a suggestion a human reviews, not an automatic action. Task ordering can be adjusted, briefs can be regenerated, and nothing is sent externally without explicit approval.
Can Changeable build something similar for our organisation?
Yes. Briefable reflects the same use case-led, governed approach Changeable applies in client engagements, including use case development, AI app development and AI governance.
Want to see what Briefable can do for your working day?
Explore Briefable directly, or talk to Changeable about applying the same modular, governed approach to your own AI product or workflow.