Pre-built briefs for the use cases we ship most.

# Scoping brief · Support deflection chatbot

## Context
We are [COMPANY], [INDUSTRY] with [N] customers. Our support team handles
[N] tickets / week, of which roughly [%] are repeat questions our docs
already answer (or should answer).

## What we want
- A customer-facing chatbot embedded in [SURFACE: in-app | help center | landing page]
- RAG over [SOURCES: docs, FAQ, policy, knowledge base]
- Citation-required so users can verify answers
- Escalation queue for questions the bot can't confidently answer
- Eval suite seeded with our 50 most common tickets

## Constraints we know about
- [DATA RESIDENCY: e.g., EU only]
- [COMPLIANCE: e.g., SOC 2 in progress]
- [INTEGRATION: e.g., must read user's subscription tier from our backend]
- [BRAND: brand voice guide attached / on our public site at LINK]

## Numbers we'd like to hit
- Deflection rate target: [%]
- Median response time: [SECONDS]
- Cost per conversation: under $[AMOUNT]
- Eval suite size at launch: 50+ cases

## Owner
[NAME, TITLE] will own this on our side post-launch. Available for
[N] hours/week during the build.

## Budget envelope
[$RANGE] for build · [$RANGE] / month for run + retainer

## Timeline
We'd like to be in production by [DATE]. Can flex by [N] weeks.

## What we've tried
[LIST: previous vendor, in-house attempt, no prior work, etc.]

## What success looks like in 90 days
[1–2 SENTENCE NARRATIVE]

# Scoping brief · Inbound lead qualification agent

## Context
We are [COMPANY], [INDUSTRY] with [N] employees. We get [N] inbound leads
per [WEEK | MONTH] from [SOURCES: website, content, ads, partners]. Our
[N] SDRs can properly follow up on roughly [N] of them.

## What we want
- Agent that engages every inbound in under 60 seconds
- Multi-channel: [CHANNELS: web chat, email reply, SMS]
- Qualification flow runs: pain, timeline, budget signal, decision authority, fit
- Qualified leads auto-booked into [AE CALENDAR: Cal.com / Calendly / HubSpot]
- Unqualified leads routed to [NURTURE: resource hub, content library, async sequence]

## ICP
We sell to [SEGMENT: e.g., B2B SaaS at 50-500 employees in NA].
Qualified = [DEFINITION: e.g., timeline ≤ 3 months + buyer involved].

## Constraints
- [CRM: HubSpot / Salesforce / Pipedrive]
- [TIMEZONE: AE coverage across XYZ]
- [LANGUAGES: support EN + ?]
- [VOICE: must match our brand guide]

## Numbers we'd like to hit
- Time-to-first-touch: < 60 seconds
- Booking rate on qualified: 80%+
- Close rate on agent-booked vs SDR-booked: within 10% delta
- No additional headcount required

## Owner
[NAME, TITLE]. Will work with [SDR/AE NAMES] on tuning.

## Budget + timeline
[$RANGE] · target launch [DATE]

## What we've tried
[Examples: HubSpot workflows, Drift, manual SDR scaling, etc.]

# Scoping brief · Document extraction pipeline

## Context
We are [COMPANY], [INDUSTRY]. We process [N] documents / [PERIOD] of type
[DOC TYPE: invoices, BOLs, customs forms, contracts, claims]. Today, [WHO]
keys them in / [HOW LONG] per doc.

## What we want
- Extraction pipeline that handles [N]% of clean docs straight-through
- Confidence-gated review queue for ambiguous extractions
- Output structured records into [TARGET: ERP, accounting system, data warehouse]
- Audit log preserving raw doc + extracted values + reasoning

## Schema
We need these fields extracted, with confidence per field:
[FIELD LIST: vendor_name, total_amount, currency, line_items, dates, ...]

## Validation rules
[BUSINESS RULES: e.g., total = sum(line_items), date within 90 days, etc.]

## Constraints
- [VOLUME: e.g., 5K docs/month]
- [LATENCY: e.g., < 5 minutes per doc]
- [PII: docs contain personal data, retention per [JURISDICTION]]
- [ON-PREM: required / optional / not needed]

## Numbers we'd like to hit
- Straight-through processing rate: 70–85%
- Field-level accuracy on STP: 99%+ for amounts and dates
- Exception queue turnaround: < 1 business day

## Owner
[NAME, TITLE]. Exception reviewers: [NAMES / TEAM].

## Budget + timeline
[$RANGE] · target launch [DATE]

## Sample documents
We will provide [N] historical documents (anonymized as needed) as the
eval-suite seed.

# Scoping brief · Multi-agent ops monitor

## Context
We run [OPERATION: plant / warehouse / fleet] with [SCALE: N machines /
N pickers / N vehicles]. Our supervisors handle [SPAN] each. Today, our
ops is reactive — issues are flagged by humans noticing them.

## What we want
- [N] specialized agents, one per role we'd hire if we had infinite humans
- Coordinating through a shared event bus
- Alerts route via [CHANNEL: WhatsApp / SMS / dashboard / Slack]
- No agent takes physical action — humans always pull the trigger
- Audit trail tying every alert back to source data

## Agent list we're thinking
- [AGENT 1: production monitor]
- [AGENT 2: maintenance scheduler]
- [AGENT 3: quality anomaly]
- [AGENT 4: inventory reorder]
- [AGENT 5: shift briefing]
- [AGENT 6: energy / cost optimizer]

## Data sources
- [SOURCE 1: SCADA, frequency, access pattern]
- [SOURCE 2: ERP, REST / direct DB / batch export]
- [SOURCE 3: operator tablets, structured inputs]
- [SOURCE 4: historical logs]

## Constraints
- [DEPLOYMENT: on-prem / hybrid / cloud]
- [LATENCY: real-time / 5-min / 15-min acceptable]
- [DATA RESIDENCY]
- [INTEGRATION: must speak to existing supervisor dashboard]

## Numbers we'd like to hit
- Unplanned downtime reduction: 20%+ over 90 days
- Mean time to detect anomaly: < 5 minutes
- Supervisor acceptance rate on alerts: 85%+
- Alert false-positive rate: < 15%

## Owner
[NAME, TITLE] — plant manager / ops director equivalent.

## Budget + timeline
[$RANGE] · phased build over [N] weeks starting [DATE]

# Scoping brief · AI strategy phase

## Context
We are [COMPANY], [INDUSTRY] with [N] employees. Leadership is being
asked [WHAT IS LEADERSHIP BEING ASKED]. We have [STATE OF AI: nothing
shipped / pilots running / one prod system / multiple]. Engineering is
[CAPACITY DESCRIPTION].

## What we want
- 12-month AI roadmap with prioritized use cases
- ROI math for top 5 candidates
- Governance framework aligned to [REGS: SOC 2 / EU AI Act / HIPAA / DPDP]
- Build-vs-buy analysis for the use cases we're considering
- Two-day engineering enablement workshop for our team

## Stakeholders to interview
[LIST: CEO, CTO, VP Eng, VP Product, VP Ops, GRC lead, ...]

## Constraints
- [BUDGET ENVELOPE FOR YEAR 1 OF AI WORK]
- [INTERNAL HIRE PLAN: yes/no/maybe]
- [DATA: cleanliness, accessibility, residency]
- [REGULATORY: what we can't cross]

## What we want to walk away with
- Written 30-page report
- Working session with our exec team
- Ranked use-case backlog with ROI estimates and complexity scores
- A first scoped sprint we can immediately green-light

## Owner
[NAME, TITLE] — typically VP Eng, CTO, or Head of AI Strategy.

## Budget + timeline
[$RANGE] · 6–8 weeks