CBS

Today’s focus (30 min): map what Copilot can draft this week vs what needs a Salesforce integration — then try a weekly-update agent (~15 min hands-on).

From pre-work

• Faster quotes — less blank-page time on recurring quote structures
• Salesforce opportunities from Outlook — creating opportunities without retyping context (plugin / integration interest)
• Early visibility on upcoming projects — knowing what is opening before the room does

Value for Sales

• Quote and proposal first drafts — structure and spec language from approved templates (you verify pricing and terms)
• Meeting prep that cites sources — research a client before the room and flag thin evidence
• Client decks on the MK template — build slides inside PowerPoint when deck prep is the bottleneck

Starting points for Sales

• Copilot in Outlook — thread summaries and reply drafts you paste into Salesforce yourself
• Copilot in Word — quote narratives and proposal sections from prior examples
• Copilot in PowerPoint — client and QBR decks from the MillerKnoll template

Honest boundaries

• Outlook → Salesforce automation is an integration decision — today we clarify handoff steps AI can shorten vs what IT must wire
• Future project visibility needs a defined data source — AI summarises what you export; it cannot invent pipeline

MK AI POV: use Copilot first for the layer around Sales work — summarising emails, drafting quote/proposal language, preparing meeting notes, and creating a cleaner Salesforce handoff. Creating or changing CRM records should remain a controlled system workflow, not a chat shortcut.

Leave with

One list: draft in Copilot this week vs needs integration or manual CRM step.

Questions for Sales

• Which step on quotes or opportunities costs the most retyping today?
• What do you already verify by hand that AI must never skip?

MK AI POV: verification for Sales means pricing, terms, availability, client commitments, and pipeline status. AI can draft the message; the account owner still owns the commercial truth.

Try it now — Weekly Update Agent (~15 min)

1. Go to microsoft365.com/chat.
2. Copy and send the simple prompt below. Wait for Copilot to reply Ready.
3. Copy and send the sample notes (Step 2) as your next message.

When I paste my messy weekly notes in the next message, turn them into a short status update for my manager.

Format:
- Highlights (max 3 bullets)
- Risks / blockers (max 3 bullets)
- Next steps (max 3 bullets)

Rules:
- Outcomes over activity
- If a number or date is not in my notes, write [Needs input] — do not invent it
- Keep it under 200 words

Do not draft the update yet. I will paste my notes in my next message. Reply "Ready — waiting for your notes." when you understand these rules.

When I paste my messy weekly notes in the next message, turn them into a short status update for my manager. Format: - Highlights (max 3 bullets) - Risks / blockers (max 3 bullets) - Next steps (max 3 bullets) Rules: - Outcomes over activity - If a number or date is not in my notes, write [Needs input] — do not invent it - Keep it under 200 words Do not draft the update yet. I will paste my notes in my next message. Reply "Ready — waiting for your notes." when you understand these rules.

weekly dump — week ending 20 Jun (messy)

=== random notes ===
- Tue call w/ dealer re: NWG — they want updated phasing options before CFO review
- pricing model v3 still w/ regional finance?? no reply yet
- Project Helios: client loved moodboards, waiting on spec feedback (email 12 Jun?)
- pipeline cleanup — moved 2 opps to Q3, one duplicate account fixed

=== email fragment (fwd) ===
From: facilities contact @ NWG
"...pilot went better than expected but we cannot commit capex until autumn planning. Can you show a phased approach that keeps floor 3 live while we plan floors 4–5?"

=== meetings (half-remembered) ===
- internal sync: marketing wants case study quote — NOT approved yet
- travel next week Amsterdam + Munich
- forgot to log SF activity on Helios until Thursday

=== blockers / risks ===
- pricing data delay (region)
- new CFO = shorter updates, hates fluff (boss said "lead with decisions needed")
- dealer asking for competitor battlecard we don't have updated

=== wins? ===
- Helios concept direction aligned
- NWG still engaged despite budget freeze

metrics: [none handy]

Make it reusable — minimal agent

Open Agent Builder on your computer (not on your phone — Agent Builder needs a desktop browser):

1. Go to microsoft365.com/chat .
2. In the left menu, click Agents, then New agent, then Skip to configure.
3. You will work on the Configure tab (fill in the fields), then Try it (test your agent), then click Create when you are ready to save.

1. Name: Weekly Update Agent
2. Description: Turns messy weekly notes into a concise manager update.
3. Instructions: copy the minimal instructions below.
4. Conversation starter: paste the sample notes and ask for this week’s update.
5. Try it with the same sample. Click Create when satisfied.

You turn messy weekly notes into a concise status update for a MillerKnoll associate's manager.

Output sections: Highlights, Risks / Blockers, Next Steps (max 3 bullets each).
Rules: outcomes over activity; mark missing facts as [Needs input]; under 200 words; no invented metrics.

On the Configure tab, paste each block from your interview into the matching box:

Keep going: full Weekly Update path (personalization interview and testing plan).

Today’s focus (30 min): try an on-brand launch-copy agent (~15 min hands-on), then agree three never-without-legal rules for imagery.

From pre-work

• On-brand copy from structured inputs — product facts → channel-ready drafts with refusal when claims are unsupported
• Imagery prompts — brief generative tools without drifting off brand
• Legal grey areas — usage rights, likeness in AI-enhanced imagery, customer data in website admin workflows

Value for Product Marketing

• Audience tailoring — adapt the same core story for dealer, architect, or facilities audiences
• Competitive and positioning frames — battlecards from approved source material only
• Rights-aware imagery briefs — review gates before assets go to design or legal

Starting points for Product Marketing

• Copilot in Word — briefs, launch narratives, and executive summaries from existing notes
• Copilot in PowerPoint — first-pass launch decks from an approved outline
• Prompt Library — copy-ready prompts for value props and objection handling

Honest boundaries

• Customer account data stays out of prompts — use anonymised summaries and internal-only drafts
• Likeness and legacy imagery need legal and brand guardrails — approval path stays human

MK AI POV: marketing agents should be grounded in approved brand voice, product claims, and channel guidance. They should flag unsupported claims instead of polishing them, and any generative imagery or likeness use needs brand/legal review before it becomes campaign material.

Leave with

One agent concept (inputs, outputs, verify step) and three never-without-legal rules for imagery.

Questions for Product Marketing

• Which launch step is still mostly manual rewrite — and what approved doc should every draft mirror?
• What fields from Airtable (or similar) are safe to ground an agent on?

MK AI POV: a safe Airtable-to-copy pattern starts with non-sensitive product facts, approved positioning, and a required review step. Do not ground a shared marketing agent on customer-identifiable records or unapproved claims.

Try it now — On-brand launch copy agent (~15 min)

1. Go to microsoft365.com/chat.
2. Copy and send the simple prompt below. Wait for Copilot to reply Ready.
3. Copy and send the sample brief (Step 2) as your next message.

When I paste a structured product brief in the next message, turn it into channel-ready launch copy:
1. Headline (max 12 words)
2. Three benefit bullets (dealer-facing)
3. One sustainability line — only if the brief includes approved facts; otherwise write [Needs input]

Rules:
- Do not invent pricing, percentages, certifications, or competitor claims
- Flag unsupported statements with [Needs verification]
- Match a professional B2B tone — no superlatives unless the brief includes them

Do not draft anything yet. I will paste my brief in my next message. Reply "Ready — waiting for your brief." when you understand these rules.

When I paste a structured product brief in the next message, turn it into channel-ready launch copy: 1. Headline (max 12 words) 2. Three benefit bullets (dealer-facing) 3. One sustainability line — only if the brief includes approved facts; otherwise write [Needs input] Rules: - Do not invent pricing, percentages, certifications, or competitor claims - Flag unsupported statements with [Needs verification] - Match a professional B2B tone — no superlatives unless the brief includes them Do not draft anything yet. I will paste my brief in my next message. Reply "Ready — waiting for your brief." when you understand these rules.

Product: CBS Task Chair refresh (fictional)
Channel: dealer one-pager + website hero
Status: pre-launch — claims NOT legal-approved yet

=== Airtable export (messy) ===
product_name: CBS Task Chair
launch_window: Autumn 2026
key_features: adjustable lumbar, 12hr sit, mesh back
sustainability_note: recycled content % TBD — ask PM
price_position: premium mid — no numbers in this draft
competitor_ref: "faster lead times" mentioned in dealer call — unverified
audience_primary: facilities managers, UK + Nordics
tone: professional, not hype

=== internal note ===
- do NOT say "best in class" — brand said no superlatives
- imagery: legacy photo may need likeness check before AI enhance
- dealer wants bullet list for email, not essay

Make it reusable — minimal agent

Open Agent Builder on your computer (not on your phone — Agent Builder needs a desktop browser):

1. Go to microsoft365.com/chat .
2. In the left menu, click Agents, then New agent, then Skip to configure.
3. You will work on the Configure tab (fill in the fields), then Try it (test your agent), then click Create when you are ready to save.

1. Name: CBS Launch Copy Agent
2. Description: Turns structured product briefs into on-brand dealer-facing copy.
3. Instructions: copy the minimal instructions below.
4. Conversation starter: paste the sample brief and ask for launch copy.
5. Try it with the same sample. Click Create when satisfied.

You turn structured product briefs into on-brand launch copy for CBS Product Marketing.

Output: Headline, three benefit bullets, optional sustainability line.
Rules: never invent pricing, metrics, or competitor claims; mark gaps as [Needs input] or [Needs verification]; professional B2B tone; no superlatives unless provided in the brief.

On the Configure tab, paste each section from Copilot's output into the matching box:

Keep going: Build Your Own Agent (full triage and testing plan).

Today’s focus (30 min): try a competitor-brief agent (~15 min hands-on), then choose one workflow to spec: competitor landscape, opportunity funnel, IP reference extraction, or sustainability data gaps.

From pre-work

• Cross-system data → summary → visualisation — competitor landscape vs CBS products, sustainability metrics
• Controlled workflows — defined inputs, consistent output format
• Opportunity funnel — turn insight, competition, anecdotal, and data evidence into concept proposals leadership can review
• Legacy IP reference file — extract key patent, design-registration, trademark, region, date, and status details from correspondence and PDFs into Excel
• Pricing support — formula checks and margin scenarios (human verifies every number)
• Tool question — when Copilot in M365 vs Claude / ChatGPT (see tool-choice frame in Concepts)

Value for Product Development

• Concept and proposal first drafts — early brief into a stage-gate template
• Opportunity-funnel summaries — consolidate inputs into a reviewable concept proposal with explicit source notes
• IP correspondence extraction — draft Excel rows from emails and PDFs, with missing fields flagged for human review
• Sustainability language with guardrails — claims only from grounded sources; flag gaps
• Excel exploration — trends and variance when data is already in a table

Starting points for Product Development

• Copilot in Excel — explore tables and summarise variance (you confirm every figure)
• Copilot in Word — concept proposals, opportunity-funnel summaries, sustainability narratives, competitor memos
• Copilot with PDFs / email text — test extraction of IP correspondence into a structured table before any automation
• Claude / ChatGPT (policy permitting) — longer external scans you verify against approved sources

Honest boundaries

• Live dashboards need data ownership — today outputs a workflow spec, not a working dashboard
• Part-level sustainability queries fail when source data is unclear — agent flags gaps, does not fabricate
• Third-party evidence and photos in opportunity-funnel work need policy review before they ground an agent or proposal
• IP data extraction can reduce admin, but Legal ownership and human review remain the control point

MK AI POV: Product Development should use AI to make uncertainty visible. The most valuable output is often not a confident answer; it is a structured proposal, source map, missing-data list, or decision brief that shows what still needs human or legal review.

Leave with

One workflow spec: inputs, output template, verification step, data owner, and data that must exist before automation.

Questions for Product Development

• Which single workflow wins if we only ship one pilot in the next 30 days?
• Where are sustainability or compliance claims hardest to draft without overstatement?
• Could IP correspondence extraction start as a reviewed Excel-helper workflow before becoming an automation?

MK AI POV: start IP and sustainability work as assisted extraction, not autonomous decision-making. Copilot can help draft rows, summaries, and issue lists; ownership, interpretation, and filing strategy stay with the responsible business and legal owners.

Try it now — Competitor landscape agent (~15 min)

1. Go to microsoft365.com/chat.
2. Copy and send the simple prompt below. Wait for Copilot to reply Ready.
3. Copy and send the sample sources (Step 2) as your next message.

When I paste competitor and product notes in the next message, produce a one-page competitor landscape brief with fixed sections:
1. Competitor overview (facts from sources only)
2. Feature comparison (our product vs competitor — mark [Needs input] where specs are missing)
3. Positioning implications for CBS (hypotheses labelled as such)
4. Gaps and verification needed

Rules:
- Cite the source for every competitor fact
- Do not invent pricing, test results, or sustainability metrics
- Label assumptions clearly

Do not draft anything yet. I will paste my sources in my next message. Reply "Ready — waiting for your sources." when you understand these rules.

When I paste competitor and product notes in the next message, produce a one-page competitor landscape brief with fixed sections: 1. Competitor overview (facts from sources only) 2. Feature comparison (our product vs competitor — mark [Needs input] where specs are missing) 3. Positioning implications for CBS (hypotheses labelled as such) 4. Gaps and verification needed Rules: - Cite the source for every competitor fact - Do not invent pricing, test results, or sustainability metrics - Label assumptions clearly Do not draft anything yet. I will paste my sources in my next message. Reply "Ready — waiting for your sources." when you understand these rules.

=== COMPETITOR SCAN (rough notes) ===
Competitor: Steelcase Series 1 refresh (fictional)
Source: public dealer spec sheet PDF (pasted excerpt)
- mesh back, 4-way arms, 135kg rating
- lead time "6–8 weeks" on sheet — verify for UK
- price point: NOT on sheet

Our product: CBS Task Chair refresh
Gap we think we have: sustainability story — but % recycled not confirmed
Gap we do NOT have: independent test data on this competitor

=== internal Teams snippet ===
PM: "Do not put competitor pricing in the brief — we never have reliable numbers."

=== sustainability fragment ===
Target: lower embodied carbon vs prior gen — no LCA number yet

Make it reusable — minimal agent

Open Agent Builder on your computer (not on your phone — Agent Builder needs a desktop browser):

1. Go to microsoft365.com/chat .
2. In the left menu, click Agents, then New agent, then Skip to configure.
3. You will work on the Configure tab (fill in the fields), then Try it (test your agent), then click Create when you are ready to save.

1. Name: CBS Competitor Brief Agent
2. Description: Produces structured competitor landscape briefs from pasted sources.
3. Instructions: copy the minimal instructions below.
4. Conversation starter: paste the sample sources and ask for a landscape brief.
5. Try it with the same sample. Click Create when satisfied.

You produce competitor landscape briefs for CBS Product Development.

Fixed sections: Competitor overview, Feature comparison, Positioning implications, Gaps and verification.
Rules: cite sources for competitor facts; mark missing specs as [Needs input]; never invent pricing or test data; label assumptions.

On the Configure tab, paste each section from Copilot's output into the matching box:

Keep going: Lifestyle Product Development workshop (concept proposals and strategy review).

Today’s focus (30 min): try a weekly commentary agent (~15 min hands-on), then name the verification step that never gets skipped.

From pre-work

• Weekly sales reports — less time rebuilding the same pack each cycle
• Order review and release process — clearer SOP language or checklist drafts
• Credit policy writing — structure from working papers, not legal finalisation in the room

Value for Finance

• Variance explanation — what moved, what needs investigation, what is noise
• Policy and process documentation — credit policy outlines and release checklists from working notes

Starting points for Finance

• Copilot in Excel — analysis on data already in a table; treat output as a hypothesis to verify
• Copilot in Word — commentary drafts and summary memos from your workbook or notes
• Copilot in Outlook — thread summaries after cross-functional reviews

Ground rules for Finance

• Every number, rate, and allocation must be traced to source — Copilot does not replace your sign-off
• Do not paste confidential financials into tools outside MillerKnoll policy

MK AI POV: Finance is a high-verification domain. Treat AI output as commentary scaffolding, not evidence. If a number cannot be traced back to the workbook, ERP extract, or approved source, it does not go into the final message.

Leave with

One named report or process for a Copilot-assisted first draft this month, plus the verification step that never gets skipped.

Questions for Finance

• Which recurring report would benefit most from a faster first draft?
• What source documents would ground a credit policy outline safely?

MK AI POV: policy drafts are a good use case when the source set is controlled. AI can propose structure, plain-language wording, and gaps to review; final policy language needs the normal Finance, Legal, and leadership approval route.

Try it now — Weekly commentary agent (~15 min)

1. Go to microsoft365.com/chat.
2. Copy and send the simple prompt below. Wait for Copilot to reply Ready.
3. Copy and send the sample notes (Step 2) as your next message.

When I paste my weekly sales notes in the next message, draft commentary for a finance review pack:
1. Headline (one sentence — what moved vs plan, if stated)
2. Key drivers (max 3 bullets)
3. Items needing investigation (max 3 bullets)
4. Suggested talking points for leadership (max 3 bullets)

Rules:
- Do not invent revenue figures, percentages, or FX impacts
- Mark missing numbers as [Needs input]
- Treat every figure as a hypothesis until the reviewer traces it to source

Do not draft anything yet. I will paste my notes in my next message. Reply "Ready — waiting for your notes." when you understand these rules.

When I paste my weekly sales notes in the next message, draft commentary for a finance review pack: 1. Headline (one sentence — what moved vs plan, if stated) 2. Key drivers (max 3 bullets) 3. Items needing investigation (max 3 bullets) 4. Suggested talking points for leadership (max 3 bullets) Rules: - Do not invent revenue figures, percentages, or FX impacts - Mark missing numbers as [Needs input] - Treat every figure as a hypothesis until the reviewer traces it to source Do not draft anything yet. I will paste my notes in my next message. Reply "Ready — waiting for your notes." when you understand these rules.

weekly sales pack notes — w/e 20 Jun (messy)

=== headline numbers (from workbook — verify) ===
UK: +3% vs plan?? need to confirm
Nordics: soft — dealer delay on one project
Total revenue: [pull from pack — not pasted here]

=== commentary fragments ===
- large order slipped from Jun to Jul (Helios account)
- FX mentioned on call — no rate in notes
- returns spike in one SKU — might be data entry error

=== what leadership wants ===
short narrative: what moved, what to watch, what needs investigation
NO invented percentages — if I did not paste a number, write [Needs input]

Make it reusable — minimal agent

Open Agent Builder on your computer (not on your phone — Agent Builder needs a desktop browser):

1. Go to microsoft365.com/chat .
2. In the left menu, click Agents, then New agent, then Skip to configure.
3. You will work on the Configure tab (fill in the fields), then Try it (test your agent), then click Create when you are ready to save.

1. Name: CBS Weekly Commentary Agent
2. Description: Drafts finance review commentary from weekly sales notes.
3. Instructions: copy the minimal instructions below.
4. Conversation starter: paste the sample notes and ask for commentary.
5. Try it with the same sample. Click Create when satisfied.

You draft weekly sales commentary for CBS Finance reviewers.

Output: Headline, Key drivers, Items needing investigation, Talking points (max 3 bullets each).
Rules: never invent figures; mark missing numbers as [Needs input]; flag items that need source verification.

On the Configure tab, paste each section from Copilot's output into the matching box:

Keep going: Copilot in Excel — data analysis (same verification discipline).

Today’s focus (30 min): try an enquiry-reply agent (~15 min hands-on), then define what stays manual until Shopify integration exists.

From pre-work

Three Internal Sales responses shared the same story — operational friction:

• Webshop enquiries — faster, consistent replies to basic questions
• Stock availability — spreadsheet not everyone checks; customers need accurate Shopify status
• Double entry — the same information typed into multiple places
• Pro formas — speeding up a repetitive raising process

Value for Operations

• Stock-update SOPs — document the real process so fewer people rely on a hidden spreadsheet
• Pro forma checklists — structure from approved examples

Starting points for Operations

• Copilot in Outlook — enquiry reply drafts for human review before sending
• Copilot in Word — SOPs for stock updates and pro forma steps
• Copilot in Excel — explore stock tables you already maintain (you verify figures)

Honest boundaries

• Shopify stock sync and auto reporting feeds are integration projects — today we define process and draft assets
• Double entry often needs system design — AI documents the handoff while ops owners pick the tool path

MK AI POV: Operations can get value before integration by standardising language, checklists, and escalation rules. For anything customer-facing, AI drafts should preserve the human review step until stock, pricing, and account data are reliably connected.

Leave with

One enquiry reply pattern or stock-update SOP outline the team can use next week.

Questions for Operations

• Which hurts most this week — enquiries, stock truth, pro formas, or double entry?
• What would “good enough to send” look like for an AI-assisted enquiry reply?

MK AI POV: define “good enough” as accurate, policy-safe, and clear about what is confirmed vs what needs escalation. AI should never invent stock, lead time, price, or account-specific terms.

Try it now — Enquiry reply agent (~15 min)

1. Go to microsoft365.com/chat.
2. Copy and send the simple prompt below. Wait for Copilot to reply Ready.
3. Copy and send the sample enquiry (Step 2) as your next message.

When I paste a webshop enquiry and internal notes in the next message, draft a customer reply with:
1. Acknowledgement
2. Stock / availability answer (only from notes — otherwise [Check stock system])
3. Next step (what you need from the customer or internal team)
4. Sign-off line

Rules:
- Do not invent stock counts, prices, or delivery dates
- Do not quote volume discounts — escalate to sales
- Keep tone helpful and professional

Do not draft anything yet. I will paste the enquiry in my next message. Reply "Ready — waiting for the enquiry." when you understand these rules.

When I paste a webshop enquiry and internal notes in the next message, draft a customer reply with: 1. Acknowledgement 2. Stock / availability answer (only from notes — otherwise [Check stock system]) 3. Next step (what you need from the customer or internal team) 4. Sign-off line Rules: - Do not invent stock counts, prices, or delivery dates - Do not quote volume discounts — escalate to sales - Keep tone helpful and professional Do not draft anything yet. I will paste the enquiry in my next message. Reply "Ready — waiting for the enquiry." when you understand these rules.

webshop enquiry — copy/paste (fictional)

From: customer@example.com
Subject: Stock on CBS Task Chair — Charcoal mesh

"Hi, is the CBS Task Chair in charcoal mesh in stock for delivery to Manchester next week? Also do you do volume discount for 40 units?"

=== internal notes (not for customer) ===
- stock sheet last updated Tuesday — charcoal mesh showing 12 units?? verify in Shopify
- volume pricing: always escalate to sales — never quote in first reply
- standard lead time if in stock: 5–7 days UK

Make it reusable — minimal agent

Open Agent Builder on your computer (not on your phone — Agent Builder needs a desktop browser):

1. Go to microsoft365.com/chat .
2. In the left menu, click Agents, then New agent, then Skip to configure.
3. You will work on the Configure tab (fill in the fields), then Try it (test your agent), then click Create when you are ready to save.

1. Name: CBS Enquiry Reply Agent
2. Description: Drafts webshop enquiry replies from pasted enquiries and internal notes.
3. Instructions: copy the minimal instructions below.
4. Conversation starter: paste the sample enquiry and ask for a reply draft.
5. Try it with the same sample. Click Create when satisfied.

You draft webshop enquiry replies for CBS Operations.

Output: Acknowledgement, Availability answer, Next step, Sign-off.
Rules: never invent stock, pricing, or delivery dates; use [Check stock system] when unknown; escalate volume pricing to sales; professional helpful tone.

On the Configure tab, paste each section from Copilot's output into the matching box:

Keep going: Weekly Update Agent (same status-rhythm pattern for ops updates).

Today’s focus (45 min): agree a team agent checklist, then choose which design pain is a near-term pilot: briefing assistant, file-finding / content review, CAD-BOM-render support, or showroom/futures simulation discovery.

From pre-work

• Best practices for building agents as a team — shared support, not solo experiments
• Visual exploration tools — Vizcom and similar alongside M365 work
• Finding information — older Teams files, information buried in email, and assets outside the local file system
• Reviewing AI-produced content — how to check sense, clarity, and source quality without losing all the time saved
• Design production tasks — CAD models, BOMs in Excel, product renders, material setup, lighting, and 3D print preparation
• Experience concepts — prompt-to-image showroom simulator and generative futures scenario modelling as longer-horizon ideas

Value for Design

• Design rationale in client language — explain choices and constraints without dumbing down the work
• Spec and finish clarity — flag missing inputs instead of inventing product details or pricing
• Visual briefs for concept work — structured prompts for exploration tools; brand review before client use
• AI-output review rubric — fast checks for sense, accuracy, source quality, and readability before a draft consumes more review time than it saved
• Production-task triage — separate what Copilot can document or structure from what CAD, render, or 3D-print tools would need to automate directly

Starting points for Design

• Copilot in PowerPoint — briefing decks from the MK template and approved example structure
• Copilot in Word — design intent memos and narrative support
• Copilot in Excel — BOM structure, missing fields, comparison against prior BOMs or quotes (human verifies supplier and material choices)
• Copilot / Microsoft Search — test search and summary workflows for Teams files and older email context before designing a larger knowledge system
• Vizcom / visual tools (policy permitting) — concept exploration; pair with morning imagery guidance

Team agent checklist (draft in session)

• Named owner and grounded source documents for each team agent
• Fixed output sections and refusal rules when inputs or specs are missing
• Internal review gate before any client-facing draft leaves CBS
• Review rubric for AI-produced content: source, sense, accuracy, audience, and action
• Quarterly refresh when templates, products, or policy change

Ground rules for Design

• Every client-facing draft is internal-first — design and account leadership review before it leaves CBS
• Product specifications and case references must come from grounded materials
• Customer floorplates, third-party imagery, material swatches, and non-MK products need policy/legal review before they are used in AI visual workflows
• CAD, render, and 3D-print automation are tool-integration questions; today can define the workflow and checklist, not automate specialist software

MK AI POV: Design should use AI as a thinking and drafting partner, not as an unreviewed creative authority. The strongest near-term pattern is controlled: approved inputs, explicit assumptions, missing-spec flags, and design review before anything reaches a client.

Leave with

Agreed team agent checklist plus one next pilot: Client Briefing Assistant, AI-output review rubric, BOM/CAD workflow spec, or showroom/futures-scenario discovery brief.

Questions for Design

• Which client deliverable has the worst blank-page problem — first briefing, revision deck, or post-visit follow-up?
• What approved decks or templates should every AI draft mirror?
• How should the team share agents — one library, one owner, or guild-style peer support?
• Which design-production task is most repetitive: CAD modelling, BOM completion, rendering setup, or 3D print preparation?
• What would a safe showroom or futures-scenario simulator need before customer data or third-party references could be used?

MK AI POV: big experience ideas are worth capturing as discovery briefs. Before a showroom simulator or futures model becomes a build, define the approved input set, customer-data rules, third-party rights, review owner, and what the experience is allowed to recommend.

Try it now — Client Briefing Assistant (~15 min)

1. Go to microsoft365.com/chat.
2. Copy and send the simple prompt below. Wait for Copilot to reply Ready.
3. Copy and send the sample brief (Step 2) as your next message.

When I paste a project brief in the next message, produce a client briefing outline:
1. Meeting purpose and desired outcome
2. Slide / section list (CBS briefing format)
3. Talking points per stakeholder type in the room
4. Gaps and [Needs input] flags (specs, finishes, pricing, lead times)

Rules:
- Do not invent product specs, pricing, or case study claims
- Flag missing inputs instead of filling them in
- Internal-first tone — this is a draft for team review

Do not draft anything yet. I will paste my brief in my next message. Reply "Ready — waiting for your brief." when you understand these rules.

When I paste a project brief in the next message, produce a client briefing outline: 1. Meeting purpose and desired outcome 2. Slide / section list (CBS briefing format) 3. Talking points per stakeholder type in the room 4. Gaps and [Needs input] flags (specs, finishes, pricing, lead times) Rules: - Do not invent product specs, pricing, or case study claims - Flag missing inputs instead of filling them in - Internal-first tone — this is a draft for team review Do not draft anything yet. I will paste my brief in my next message. Reply "Ready — waiting for your brief." when you understand these rules.

project brief — rough (fictional client: Meridian Offices)

client: Meridian Offices — workplace refresh, 3 floors
meeting: client briefing, 8 Jul, London
in room (them): facilities lead, procurement
in room (us): CBS design lead, account manager

objective: present concept direction for floor 2 — not final spec
deliverable: briefing deck outline + talking points

=== what we know ===
- preference for warm neutrals, not clinical white
- budget band discussed informally — no number in writing
- lead time sensitive — autumn occupancy target

=== MISSING ===
- approved case study for similar scale?
- exact finish codes for hero product
- pricing / phasing — do not invent

Make it reusable — minimal agent

Open Agent Builder on your computer (not on your phone — Agent Builder needs a desktop browser):

1. Go to microsoft365.com/chat .
2. In the left menu, click Agents, then New agent, then Skip to configure.
3. You will work on the Configure tab (fill in the fields), then Try it (test your agent), then click Create when you are ready to save.

1. Name: CBS Client Briefing Assistant
2. Description: Produces client briefing outlines from project briefs.
3. Instructions: copy the minimal instructions below.
4. Conversation starter: paste the sample brief and ask for a briefing outline.
5. Try it with the same sample. Click Create when satisfied.

You produce client briefing outlines for CBS Design.

Output: Meeting purpose, Section list, Talking points by stakeholder, Gaps and flags.
Rules: never invent specs, pricing, or case studies; mark missing inputs as [Needs input]; internal review before client-facing use.

On the Configure tab, paste each section from Copilot's output into the matching box:

Keep going: Client research & meeting prep (full meeting-prep agent with grounding docs).