A scope, written in two minutes. Not six hours.
Case studyScope AssistantVideo 1:26
We built a Claude-powered scoping assistant for BWP Group. It reads a client call transcript, matches it against past scopes and price benchmarks, runs a 22-prompt framework and saves a formatted scope to Google Drive. Scoping that took about six hours end to end now produces a first draft in two minutes. Live since March 2026 after a five-week build, it has produced more than 200 proposals, and under-scoped projects are down 22%.
- Client
- BWP Group
- Category
- Operations
- Engagement
- 5-week build, live since March 2026, results to October 2026
- Stack
- Claude, Fireflies or Gong transcripts, Google Drive or SharePoint, the client’s SOW library and price benchmarks
- Time to value
- 5 weeks, build and training
- Investment
- £7,000
- Status
- Live, 200+ proposals produced
- Updated
Results
Measured from go-live in March 2026 to October 2026.
| Metric | Before | After | Measured over | How measured |
|---|---|---|---|---|
| Time to a scope | About 6 hours, end to end | First draft in about 2 minutes | March to October 2026 | Before: BWP Group’s own figure. After: the assistant’s run time. |
| Proposals produced | None (new system) | 200+ | March to October 2026 | Scopes generated by the assistant. |
| Under-scoped projects | Before go-live | 22% fewer | March to October 2026 | Compared with projects scoped before go-live. |
See it in action
86 seconds. Start with a client call transcript, finish with a scoping document in Drive.
Read the transcript
From Vimeo’s auto-generated captions, lightly edited.
What it does
So what this tool does is it creates, ultimately, scopes. So loads of our clients, or us, we get calls saying, “Please, can you do this?” So instead of it taking hours for someone to put together a scope, jot down notes, instead, what we do is we take a transcript.
What goes in
You put in any information that you may have that’s specific that won’t be covered in that transcript. You put in the budget you discussed, and you put in anything else. Here on then specify use Core Signals API.
Generating the scope
You click Generate Scope, and then, based off historics of the best scopes that have had the highest acceptance rate, what it does, it goes and creates that scope. So it will analyse all the previous ones. So there’s a 15-step prompt approach to each individual section. It will chunk up the input documents, understand it in detail, and then what it’ll do is it’ll create a scope document exactly how I want it to. So now it’s going, it’s chunking, it’s understanding. It may take a minute or so, but it’s just putting in the specifics, then it generates a document.
The document in Drive
This can be generated and put either in Microsoft or on Google Drive. I use Drive, so it exclusively goes here. You can see my logo. You can see the header. You can see an overview section, core goal, KPIs, technological approach, inputs, outputs, considerations, scope, total cost, deliverables, and timeline.
The result
So it’s essentially taken a task that used to take the team hours and now done in minutes. And this can be done for any client and customised to their tone of voice and how they like the scope.
What it does
The Scope Assistant turns a client conversation into a finished, branded scoping document, automatically. It pulls from your historic SOW library, your price benchmarks, and a 22-prompt framework that mirrors how your best people scope.
Reads client calls
Drop a transcript in. The assistant pulls out scope shape, constraints, budget and stakeholder language.
Knows your past work
Trained on your historic frameworks and price benchmarks, so every new scope is anchored to real engagements.
A 22-prompt framework
Every scope document gets the same depth. Deliverables, dependencies, assumptions and exclusions: no section forgotten.
Lands in Drive
The final document uploads to Drive or SharePoint in your approved format. No copy-paste, no manual filing.
The workflow
Six stages, from the client call to a document ready to review.
-
Client call
Discovery or scoping conversation.
What happens at this step
- Inputs
- Live client conversation captured on a transcription tool.
- Outputs
- A raw transcript ready to feed the system.
- Time
- Length of the call, typically 30 to 60 minutes.
The starting point is a real conversation: a discovery call, a scoping intro, an account review. Whoever runs it captures it on Fireflies, Gong or any tool that produces a transcript. That recording becomes the system’s first input.
-
Transcript and context in
Transcript, budget, project notes.
What happens at this step
- Inputs
- Call transcript, discussed budget, project context notes.
- Outputs
- A structured intake package ready for analysis.
- Time
- Under 60 seconds.
The team uploads the transcript along with the discussed budget, project type and any contextual notes that didn’t make it onto the call. This is the only manual step in the whole workflow, and it takes under a minute.
-
Intelligence ingest
Four-source comprehension.
What happens at this step
- Inputs
- Intake package, historic frameworks, price benchmarks, requirements.
- Outputs
- A structured analysis ready for the prompt framework.
- Time
- Under 30 seconds.
The intelligence layer parses the intake against four reference sources at once: the call transcript, your historic scope framework library, internal price benchmarks and the explicit requirements stated on the call. Nothing is forgotten, nothing is improvised.
-
22-prompt analysis
The framework runs against every section.
What happens at this step
- Inputs
- Structured analysis from the intelligence layer.
- Outputs
- Section-by-section scope content, fully written.
- Time
- 45 seconds.
The agent runs the analysis through a 22-prompt framework. Each prompt maps to a section of your scope document: deliverables, dependencies, assumptions, exclusions, milestones, pricing logic, risks, success criteria. Every scope gets the same depth, every time.
-
Document assembled
Branded format, every time.
What happens at this step
- Inputs
- Section-by-section scope content and the approved template.
- Outputs
- A polished, branded scope document.
- Time
- 15 seconds.
Scope content is assembled into your approved template. Same headings, same tone of voice, same formatting, same branding. Every document leaving the system reads like one a senior lead would write by hand.
-
Upload to Drive
Ready to review and share.
What happens at this step
- Inputs
- Finished scope document.
- Outputs
- A reviewed, shareable document in your storage of record.
- Time
- Instant.
The finished document is uploaded to the correct folder in Drive or SharePoint, named to your convention, and the reviewer is notified. The senior lead opens it, edits if needed, and ships it to the client.
The problem
Every proposal was hand-written from scratch. Scoping quality lived in one person’s head, and growth lived in their calendar.
The most experienced lead blocked out most of a day to scope each new engagement: asking the right questions, pricing against similar past work and flagging the risks they remembered. When they were away, scoping went to whoever was free, and quality dropped.
The knowledge wasn’t missing. It sat in one person’s head, in years of past SOWs in a folder no one searched, and in pricing decisions that had never been written down. Every scope was a one-off, so the team re-learned the same lessons every quarter.
- About six hours per scope, end to end, owned by a single senior lead.
- Projects regularly under-scoped against what delivery actually took.
- Past pricing and risk lessons not written down anywhere the team could use.
The solution
An intelligence system that scopes the way your best person scopes. Trained on the past, framed by 22 prompts, finished in two minutes.
An assistant that scopes like your senior lead
The Scope Assistant takes a client call transcript, contextual notes and the discussed budget. It cross-references your historic scope frameworks and price benchmarks, runs 22 prompts mapped to every section of your scoping template, and delivers a formatted document into Drive, in your house style, in two minutes flat.
- Transcript comprehensionPulls scope signals, constraints and budget from any client call recording.
- Historic framework libraryCross-references the closest comparable past engagements for shape and pricing.
- 22-prompt engineEvery scope section gets its own prompt. Nothing missed, nothing improvised.
- Auto-format and uploadLands the document in Drive or SharePoint in your approved branded template.
- Objectives
- Deliverables
- Dependencies
- Assumptions
- Exclusions
- Timeline
- Pricing
- Risks
Want a scoping assistant for your team?
Share your last ten SOWs. We will build the custom assistant in a single sprint.
Start a sprintThe outcome
Scoping went from a day of senior time to a two-minute first draft that the senior lead reviews and sends.
- 2minTo a first draftDown from about six hours end to end.
- 200+Proposals producedSince go-live in March 2026.
- 22%Fewer under-scoped projectsCompared with before go-live.
What we would do differently
What we would change if we built it again.
Start with an ontology
Agree the team’s naming conventions for projects, services and deliverables before the build, so the assistant and the people use the same words from day one.
Record the process in more detail first
Use Atlas, our process recording tool, to capture how the team actually scopes, step by step, and build from that recording.
Questions
How long does it take to build an AI scoping assistant?
BWP Group’s took five weeks, including training the team. It went live in March 2026.
What does an AI scoping assistant cost?
This build cost £7,000. Get in touch for a quote based on your own scoping process and documents.
What tools does it need?
Claude, a call transcription tool such as Fireflies or Gong, Google Drive or SharePoint, and your library of past scopes and price benchmarks.
How much has it been used?
More than 200 proposals since March 2026. Under-scoped projects are down 22% compared with before it went live.
More case studies
All our work-

Next Gen Intelligence: raw signal to brief-ready concepts
Case studyIntelligenceVideo 2:40
60,000 qualitative documents turned into trends, platforms, personas and cited creative concepts.
-

Three days of LLM SEO, in one hour
Case studyWorkflowsVideo 3:10
Keyword research, query generation and a semantic audit, chained into one automated run.
-
Weekly check-inClient callSend the revised media planBook the creative reviewRecap and next steps
Team Agent Assistants: the admin already handled
Case studyOperations
Reads each meeting transcript, reconciles tasks in Trello and drafts the client follow-up in your voice.
Scope like your best person, every time
Send us your past proposals. We will build an assistant that scopes new work in your voice, with your pricing and your risk knowledge.