Session report
You use AI most days, but many of you still double-check every finance number by hand.
Seating Group Agent Builder · 27 August 2026 · facilitated by Harvey Schaefer
Eight of you from seating product and portfolio roles replied before tomorrow’s session. You’re already using Copilot regularly, and almost everyone named the same monthly rhythm: pull the International PL, work it in Excel, rebuild the deck. What’s missing isn’t access or motivation. It’s confidence that the numbers are right, and prompts that match how you actually work.
- People who replied
- 8
- Your group, August 2026
- Use AI daily
- 5 of 8
- Most of you already have a habit
- Care about reporting
- Everyone
- The shared monthly spine
- Worry about accuracy
- Half the room
- More common here than MillerKnoll overall
Where the room actually landed
Five themes from your pre-work, with MillerKnoll and industry context where we have it.
Uses AI at least weekly
87.5%of this room, against a MillerKnoll average of 63.4% and an industry benchmark of 30%
Ready
Uses AI at least weekly
Ready
87.5% of this room; against a MillerKnoll average of 63.4% and an industry benchmark of 30%Uses AI at least weekly
87.5%
of this room; Ready; against a MillerKnoll average of 63.4% and an industry benchmark of 30%Uses AI at least weekly
87.5% of this room; Ready; against a MillerKnoll average of 63.4% and an industry benchmark of 30%Not ready below 60% · Partial 60–85% · Ready above 85%
Uses AI daily
62.5%of this room, against a MillerKnoll average of 26% and an industry benchmark of 15%
Partial
Uses AI daily
Partial
62.5% of this room; against a MillerKnoll average of 26% and an industry benchmark of 15%Uses AI daily
62.5%
of this room; Partial; against a MillerKnoll average of 26% and an industry benchmark of 15%Uses AI daily
62.5% of this room; Partial; against a MillerKnoll average of 26% and an industry benchmark of 15%Not ready below 60% · Partial 60–85% · Ready above 85%
Worried about getting numbers wrong
50%of this room, against a MillerKnoll average of 21.2% and an industry benchmark of 35%
High concern
Worried about getting numbers wrong
High concern
50% of this room; against a MillerKnoll average of 21.2% and an industry benchmark of 35%Worried about getting numbers wrong
50%
of this room; High concern; against a MillerKnoll average of 21.2% and an industry benchmark of 35%Worried about getting numbers wrong
50% of this room; High concern; against a MillerKnoll average of 21.2% and an industry benchmark of 35%Low concern below 25% · Elevated 25–40% · High concern above 40%
What this means
Half of you named accuracy or trust as a blocker, more than we usually see company-wide. Even daily users said they’re not sure what to ask when the Excel extract is messy. Tomorrow we’ll build in a simple check: every figure traces back to your file before you share anything.
Wants help rebuilding decks
50%of this room, against a MillerKnoll average of 19.2% and an industry benchmark of 25%
Not ready
Wants help rebuilding decks
Not ready
50% of this room; against a MillerKnoll average of 19.2% and an industry benchmark of 25%Wants help rebuilding decks
50%
of this room; Not ready; against a MillerKnoll average of 19.2% and an industry benchmark of 25%Wants help rebuilding decks
50% of this room; Not ready; against a MillerKnoll average of 19.2% and an industry benchmark of 25%Not ready below 60% · Partial 60–85% · Ready above 85%
What this means
Four of you called out the blank-slide problem explicitly: turning verified Excel bullets into a deck, not starting from an empty PowerPoint. That’s a bigger pain here than in typical company-wide training, which tends to underweight your monthly rebuild.
Wants help with reporting and analysis
100%of this room, against a MillerKnoll average of 36.5% and an industry benchmark of 30%
Ready
Wants help with reporting and analysis
Ready
100% of this room; against a MillerKnoll average of 36.5% and an industry benchmark of 30%Wants help with reporting and analysis
100%
of this room; Ready; against a MillerKnoll average of 36.5% and an industry benchmark of 30%Wants help with reporting and analysis
100% of this room; Ready; against a MillerKnoll average of 36.5% and an industry benchmark of 30%Not ready below 60% · Partial 60–85% · Ready above 85%
What we heard
You’re not starting from zero. Most of you use AI at least weekly, many daily, well ahead of typical MillerKnoll and external panels. Nobody framed this as a CRM or inbox problem. This is a reporting-and-story room: International PL, Excel, PowerPoint, repeat.
The friction is trust on finance outputs, not tool access. Half of you worry about accuracy on important work. That showed up again in open answers: “Do the formatting for me. I’ll own the takeaways,” but only if the numbers trace to the sheet. Deck rebuilds matter too; presentations ranked much higher for you than in company-wide surveys.
Tomorrow we focus on small, repeatable wins: one pass in Excel on a real PL extract, a slide outline from bullets you already trust, a competitive scan that labels what’s verified vs. what still needs checking, and one agent build so the monthly cycle doesn’t restart from a blank prompt.
Visual summary
How this was measuredPre-work Airtable survey, August 2026. Eight seating product and portfolio leaders in Europe. Bars show how many people named each theme. With a group this size, one person shifts a bar noticeably; treat small gaps as directional, not precise.MillerKnoll comparisons come from the enablement pre-training dashboard (266 respondents). Industry use-frequency references use Gallup Workplace Q2 2026. The trust comparison uses a directional industry proxy from Microsoft Work Trend Index 2026, not a direct Gallup cut.
What now
Four commitments. Under each one: how to do it in Copilot Chat, in Excel, in PowerPoint, or with an agent — pick the surface that matches the work.
- ChatDraft and rewrite in the Copilot pane
- ExcelTables, pairing, and number comparisons
- PowerPointShort briefs you can present
- AgentReusable checker or follow-up helper
Run a first-pass read on your PL in Excel
Open a cleaned monthly International PL or Tableau export. Ask Copilot for structure and flags, not invented totals.
Starts from the work everyone named, and builds the verification habit before numbers leave the sheet.
Do it with Copilot — Chat, Excel, PowerPoint, or Agent
- Excel
Summarize sales and margin by product and region for [July 2026]. Show MoM and YoY only where both periods exist in this sheet. List any nulls, conflicts, or rows you cannot calculate. Do not invent numbers. Output: (1) a tight table, (2) five headline bullets I can paste into PowerPoint.
- Excel
Build a slide outline from bullets you trust
Paste your Excel headline bullets into PowerPoint Copilot. Ask for titles and talk track. Mark [Needs input] anywhere data is thin.
Directly addresses the deck-rebuild pain half of you named, without starting from a blank slide.
Do it with Copilot — Chat, Excel, PowerPoint, or Agent
- PowerPoint
I am rebuilding our monthly seating performance deck for [sales leadership / Europe portfolio review]. Here are the only facts I trust from Excel. Use these exactly, do not add numbers: [paste the 5 bullets from Excel] Create an 8-slide outline: slide title, one-line takeaway, speaker note on what I still need to verify. If a slide needs data I did not provide, write [Needs input].
- PowerPoint
Run a competitive scan with confidence labels
Attach two or three real competitor sources in Copilot Chat. Ask for Verified / Needs check / Unknown on every cell.
Covers research and opportunity work without the polished fiction several of you warned about.
Do it with Copilot — Chat, Excel, PowerPoint, or Agent
- Chat
We are comparing task seating in the €800 to €1200 band for European corporate. Using ONLY the sources I attached, build a competitor matrix with columns: Product | Brand | Key features | Price band | Source | Confidence (Verified / Needs check / Unknown). Then: 5 implications for product requirements (Verified rows only) and a research backlog. Do not invent list prices or specs.
- Chat
Build the Product Performance Storyteller agent
One Agent Builder cycle on a scrubbed monthly extract so the PL → brief → deck path repeats without blank-page prompting.
Turns a habit many of you already have into a monthly workflow you can refine instead of restart.
Do it with Copilot — Chat, Excel, PowerPoint, or Agent
- Agent
Build a Copilot agent for MillerKnoll Seating PMs that turns monthly Tableau / International PL / Excel extracts into a structured performance brief: headline bullets, product/region tables, MoM/YoY movers, data-quality flags, and a slide outline. Never invent financials. Flag conflicts between sources. Use [Needs input] when data is missing. Test prompt: Here is this month's International PL extract for [product or portfolio]: [paste or attach]. Produce the full brief for [period]. Circle every figure against the file before I share.
- Agent
You’re already ahead on usage. Tomorrow isn’t about convincing anyone to try AI. It’s about building a verification habit on finance outputs and capturing your monthly reporting spine in something reusable. That’s a working session together, not a lecture.