Section 2
What we heard
Section 3
Executive summary
Customer acquisition is the biggest single lever, and the call surfaced a tight set of opportunities around it. They break cleanly into two groups: one urgent project we'd push on first, and four quick wins your team can start this week without us.
Top priority · Urgent & important
Proposal generation: distributor email to ERP, in one step
Highest impact, higher effort. Reads inbound proposal requests, matches them against your catalogue, drafts the quote from your past proposals, and writes the won deal back into the ERP. The single workflow we'd push hardest on.
Jump to Area 1 in Section 6Start this week · Four quick wins
Things you can begin without us
- AI meeting notetaker for handoffs.
- AI-assisted proposal drafting (manual version of the top priority).
- AI-assisted report summaries.
- Sanctioned AI accounts for the team.
Behind both groups, Essedi is at an inflection point with AI itself. Several people already use it on their own, and the company is moving toward official adoption. That transition matters beyond productivity: consumer AI accounts don't carry the data protections of enterprise tiers, so formalising use also closes a governance gap most teams don't notice until something goes wrong.
Based on the call, Essedi is at the transition from individual AI use to structured team adoption. This is the point where the productivity gains from individual experimentation either get formalised into something the whole business benefits from, or stay trapped in the people who found them. Most companies at this stage have the right instinct (official tools, written policy) but underestimate how much the workflow design matters relative to the tool choice.
The tools are not the hard part. What separates businesses where AI changes results from ones where it stays a side experiment is whether a specific process changes alongside the tool. The two recommendations above are organised that way: one bigger workflow change worth doing carefully, and four lower-risk pilots to start now.
Section 4
Impact-effort matrix
Each opportunity from the call, plotted by how much effort it takes to get started against how much impact it's likely to have. Position is relative, not exact. The single ringed point is the highest priority.
Tap any point for a preview, then open the full card.
Quick wins (left half, top) are detailed in Section 5. Major projects (right half, top) are covered in Section 6. Vertical position within each group reflects relative impact; horizontal position reflects relative effort.
Section 5
Quick wins
Four things you can start this week without us. Each card shows why it fits, what to do on day one, and how you'll know it's working by the end of week one.
Quick win 1 · Meeting capture
AI meeting notetaker
"Scaling from individual contributions to something more structured where teams know how to hand over tasks to each other efficiently."
What to use. Fireflies.ai is the strongest cross-platform default with reliable Italian transcription. If Essedi is on Google Workspace, the built-in Gemini in Meet is a zero-install starting point. On Microsoft 365, Copilot in Teams plays the same role. Granola is worth a look for Mac-heavy users, with a slicker UX that takes notes from your own audio instead of joining the call as a bot.
Why it matters. Two compounding benefits, only one of which most teams notice. The obvious one is documented handoffs. The less obvious one is that every transcribed meeting becomes raw material, for capturing what key people know (Area 4), for sharpening proposals (Quick win 2), and for building Essedi's institutional memory as the team grows.
Why this fits
- Sales-to-designer-to-logistics handoffs break when meeting outcomes aren't captured in writing.
- Six to seven meetings per week across functions with different vocabularies. Someone always misses what was decided.
- Italian transcription quality is the make-or-break factor. Fireflies and Granola handle it well at scale. Gemini and Copilot are built into the meeting app you already use.
- Transcripts compound: the longer the team uses one tool, the more it's worth as a knowledge base, not just a notetaker.
How to start this week
- Pick the tool that matches your meeting platform. Google Meet means Gemini, Teams means Copilot, mixed setups mean Fireflies or Granola.
- Sign up for the free tier, connect the calendar (20 minutes).
- Pick one cross-team meeting this week. A project handover or a sales-to-design brief.
- After the meeting, read the summary. Does it capture who's responsible for what?
- If yes, roll out to the next meeting type. If not, try one more. Quality improves with context.
Quick win 2 · Proposal drafting
AI-assisted proposal drafting
"We're starting to build these templates, but it's still very manual."
What to use. Claude.ai is the strongest fit for long-form drafting and document handling, with full Italian interface. ChatGPT and Gemini work the same way for this. Pick one and have the team use it consistently. The mechanism matters more than the brand: a persistent project library plus iterative drafting in plain Italian.
Why it matters. Cuts the "from scratch" problem on every new proposal, and quietly builds a house-style library as a side effect. Each new draft gets smarter without anyone managing a template, the past proposals you upload do the work. This is also the natural first step toward the connected pipeline in Area 1.
Why this fits
- AV/UC proposals are hard to template. Every job is a different mix of rooms, systems, and constraints.
- Claude's Projects feature lets you upload past proposals as a persistent library, so every new session starts from there. ChatGPT GPTs and Gemini Gems do the same.
- Describe the client's space in plain Italian. The model drafts the structure, and your team's domain expertise stays where it should.
- Full Italian interface across all three tools. Usable by systems design and sales support, not just English-comfortable staff.
How to start this week
- Pick one tool for the team and stick with it ($20/mo per seat for any of them).
- Create a Project, GPT, or Gem called "Essedi Proposals".
- Upload 2 to 3 existing proposals (different project types) plus a one-page product/service summary.
- Describe a current client opportunity in Italian and ask the model to draft a proposal structure.
- Before saving the prompt as your template, paste it into promptoptimizer.tools. Free, no signup, sharpens the instruction in seconds.
- Goal in week one: see how much of your standard vocabulary and format comes out on its own.
Quick win 3 · Reporting
AI-assisted report summaries
Asked whether anyone pulls data together or builds reports manually, including billing and utilization. Toby said yes to both.
What to use. Whichever LLM you picked for Quick win 2 (Claude, ChatGPT, or Gemini). Same account, same monthly fee. The tool barely matters here, the saved prompt does.
Why it matters. The recurring report shifts from assembly to review. The work goes from "pull, reformat, write the summary" to "paste, read, ship". The same prompt-and-paste pattern then scales to any other manual report someone produces each month.
Why this fits
- ERP plus Excel plus manual billing tracking means someone in admin pulls data, reformats it, and produces a summary someone else reads.
- The fix doesn't require touching the ERP. Paste raw Excel data into the chat, describe the output format, get the summary.
- The work shifts from assembly to review.
- Same account as Quick win 2. No additional cost.
How to start this week
- Pick the report that takes the most time to assemble. Likely billing or utilization.
- Copy the raw data out of Excel or the ERP as is. No formatting needed.
- Describe in the chat what the report should show and who reads it.
- Spend 20 minutes refining the prompt until the output matches your format. Paste it into promptoptimizer.tools to sharpen it.
- Save that prompt. From then on it's a paste-and-review job.
Quick win 4 · Governance baseline
Move the team onto sanctioned AI accounts
"It started as people doing it on their own, and now it's migrating to a more official usage."
What to use. Claude Team, ChatGPT Business, or Mistral Business. Pick one for the whole company. If Quick wins 2 and 3 landed on Claude, Claude Team is the natural extension. The decision matters less than the consolidation: one tool, one billing line, one policy.
Why it matters. The team is already using AI on personal accounts. The question isn't whether AI is in the building, it's whether what they paste is protected. One company subscription closes the data-governance gap with a single decision, and gives Essedi a defensible posture if a client ever asks how AI is governed in your operations.
Why this fits
- Personal-tier AI accounts may log prompts to vendor training datasets by default. Enterprise tiers don't.
- The team is already using AI. The choice is whether the data they paste is protected, not whether it's used.
- Italian-market relevant: the EU AI Act applies risk-tier obligations for businesses deploying AI tools in operations. A sanctioned baseline is the prerequisite for any compliance posture.
- One Team subscription covers everyone in Sales, Sales Support, Systems Design, and Marketing. Same account, same rules.
How to start this week
- Pick one platform for the company. Claude Team is the natural fit if Quick wins 2 and 3 land.
- Write a 2-page acceptable-use note: what people can paste, what they can't (client financials, employee data, signed NDAs), and which accounts to use.
- Move every existing personal account onto the company plan in one week.
- Audit shadow accounts at the end of the month. Anyone still on a personal tier gets migrated.
- Review the EU AI Act risk classification once. Most off-the-shelf use sits in "minimal risk", but the review itself is the artefact a future client audit will ask for.
Section 6
Areas worth a closer look
Proposal generation: from distributor email to ERP in one step
"Emails arrive with all sorts of attachments, someone has to open them, go into the ERP, go into Excel, and for each proposal you can lose a day."
What this could look like
- An email arrives from a distributor or end client. Free-form, with specs, room lists, or equipment references in various formats. The system reads it and extracts what matters: client, project brief, room configuration, product references.
- Each extracted line item is matched against Essedi's product catalogue and current pricing. The system looks up the right SKU and price, so the systems designer isn't cross-referencing pricelists manually.
- That matched, priced data pre-fills a draft proposal using Essedi's own past proposals as the template. The systems designer reviews and adjusts, instead of building from scratch.
- On approval, the relevant job data flows into the ERP automatically. No manual re-entry, no transcription step between winning the deal and opening the project.
Why this is harder than it looks, and more valuable
AV/UC proposals don't come in a standard format. Specs arrive as PDFs, room layouts, email text, or references to equipment configurations that need to be matched against Essedi's product range. The manual step isn't just typing. It's interpretation: reading what a client means and translating it into a structured quote. That's exactly where the time goes, and exactly where a well-designed workflow saves the most. The first step is always scoped as extraction only: read the input, produce a clean structured draft. Pricing and ERP write-back follow once the extraction is reliable.
What we'd want to understand better
- What does a typical inbound proposal request look like? Free-form email, a formal RFQ document, or both?
- How many proposals does Essedi handle per month, and how long does the current drafting step take on a typical one?
- Where does pricing live today? A pricelist in the ERP, a separate Excel, or a mix? Is there one pricelist or different ones per client or distributor?
- When a deal is won, what's the first step someone takes in the ERP to open the project, and how much of what's in the proposal gets re-entered manually?
Connects to Quick win 2 (manual, AI-assisted version). The pipeline is the next step once the manual workflow proves the format.
Outbound lead-gen for the sales team
"It's acquiring new customers. And that takes up a lot of time with the salespeople." Sales pegged at 25 to 30% of their time on acquisition. The biggest single drain named on the call.
What this could look like
- AI-assisted prospecting: identifying target companies that match Essedi's existing client profile, and surfacing contact information at the right level.
- Automated outreach sequences with personalised messaging at scale, so sales focuses on conversations rather than finding people to have them with.
- Lead scoring that prioritises inbound inquiries by fit, so the team knows where to spend time first.
What we'd want to understand better
- How does Essedi currently find new prospects? Referrals, events, cold outreach, inbound from the website?
- What does the 25 to 30% of time actually look like day-to-day? Research, outreach, follow-up, or all three?
- Is there a CRM, or is pipeline tracked in Excel and email?
Sales handoffs and ERP utilization
"Scaling from a type of operations which was based on individual contributions to something more structured where teams know how to interact properly and hand over tasks to each other efficiently."
What this could look like
- Mapping the current sales-to-delivery handoff to find where things stall, get duplicated, or fall through the cracks between sales, systems designers, and logistics.
- AI-assisted process documentation that turns what individual contributors know into written playbooks the team can follow.
- Simple workflow automation that moves information between the ERP, email, and the people who need it, without requiring anyone to chase it.
- A targeted ERP utilization audit: where the "Business" ERP is genuinely the right system of record, where Excel has taken over because the ERP can't, and where the gap could close with a thin AI layer rather than a system replacement.
A pattern worth naming
What you described, significant Excel use alongside an ERP that isn't used to its full potential, is the most common pattern in B2B companies at this stage. Work happens in email and spreadsheets, and humans manually bridge the gap into the system of record. The ERP ends up as a compliance layer rather than an operating layer. Every time someone chases a status update, copies a number from email into a system, or reconstructs a project view that should already exist, that's time that should be available for higher-value work. The opportunity isn't replacing the ERP. It's closing the gap between where decisions actually happen and where data currently sits.
What we'd want to understand better
- Where exactly do handoffs between sales, systems designers, and logistics currently happen? Is there a defined step, or does it vary by project?
- What does a failed handoff look like for Essedi? A delayed project, a missed spec, something else?
- How much of the ERP is currently used for project tracking compared to just finance?
Knowledge system for key-person dependency
Asked whether anything in the business essentially only works because one specific person knows how to do it. Toby said yes. No detail surfaced on who or what area, but the signal was clear.
Why this is harder than it looks
The knowledge that creates key-person dependency is almost never the knowledge that's already written down. It's the exceptions. The judgment calls, the client-specific workarounds, the "when we install in environments like this, here's what actually happens", that live entirely in someone's experience. Experienced people in technical companies effectively become the integration layer between systems that don't talk to each other and processes that were never fully documented. Capturing that knowledge isn't just a business-continuity move. At 55 people and growing, it becomes a scaling prerequisite: new hires ramp faster when the real operational knowledge is accessible, instead of having to learn it from the one person who carries it.
What this could look like
- A structured knowledge capture process: turning what key people know into searchable internal documentation, starting with exceptions and edge cases, not generic SOPs.
- A custom knowledge base that the team can query in natural language, so the expertise is accessible even when the person isn't available.
- Onboarding material generation from existing documentation, so new people ramp faster.
What we'd want to understand better
- What area does the key-person dependency sit in? Technical installation knowledge, client relationships, system design, something else?
- Is there any existing documentation (even informal), or is the knowledge purely in someone's head?
- How often does this dependency actually create a bottleneck compared to being more of a latent risk?
Section 7
Your next steps
