Deep Dive: AI prompts to reduce weak underwriting
Key takeaways
LLMs are most useful when you give them structured, context‑rich prompts tied to specific ETA workflows
The highest leverage shows up in three places: picking the right lane, building a target universe, and running the first owner calls that surface fragility
The goal is to compress low‑value thinking and drafting, so you can spend more time with owners and underwrite downside
Mechanics
Good prompts do three things: they constrain scope, force uncertainty, and output artifacts. The investor-grade rule is to treat every claim as unproven until it has a benchmark, a falsifier, or a document behind it.
One extra control: run the same prompt through two or three different models (e.g., ChatGPT, Gemini, Perplexity) and treat disagreements as a signal of where your inputs are thin or your assumptions are doing the work. If you want a low-friction way to do that, multi-model “wrappers” like Abacus. AI’s ChatLLM bundle access to multiple LLMs under one interface (Basic plan is listed at $10/month).
3 prompt packets (copy/paste)
1) Lane / niche identification
Use when: you’re choosing where to spend the next 4-6 weeks and want to avoid lanes that look attractive but don’t survive leverage.
Prompt
You are an ETA investor screening a vertical or an acquisition lane in Europe. Be skeptical and commercially grounded. Important warnings: do not rely on marketing language; use external benchmarks where possible; separate Facts / Assumptions / Inferences / Open Questions; flag speculation clearly. Before starting on the task, ask me up to 10 high impact questions that will help sharpen your response.
Candidate lane: [one sentence on niche + geography]
My ETA constraints: B2B-heavy, repeat/recurring preferred, low capex, strong cash conversion, modest complexity for a first-time CEO.
Tasks:
Translate the lane into 7 hard screening rules, explicitly covering: B2B vs B2C, recurring vs project, capex intensity, working-capital profile, customer concentration risk, regulation/claims risk, and key-person dependence.
Give 5 sub-niches to prioritise and 5 to avoid, with one-line reasons tied to underwriting (margin durability, cash conversion, fragility).
Provide 10 search keywords + 8 exclusion keywords that reflect the screening rules.
Output a Lane Scorecard table with 6 rows: Revenue quality / Pricing power / Capex / Working capital / Cyclicality / Operational complexity. Score 1-5 + one sentence each.
Output: concise bullets + one table.
➡️ Example input (paste as-is)
Prompt
You are an ETA investor screening an acquisition lane in Europe. Be skeptical and commercially grounded.
Important warnings: don’t rely on marketing claims; use external benchmarks where possible; separate Facts / Assumptions / Open Questions; flag speculation.
Before starting on the task, ask me up to 10 high impact questions that will help sharpen your response.
Candidate lane: Fire protection inspection + maintenance for commercial buildings in Spain and Portugal (alarms, sprinklers, extinguishers, compliance inspections).
My ETA constraints: B2B-heavy, repeat/recurring preferred, low capex, strong cash conversion, modest complexity for a first-time CEO.
Tasks:
Translate this lane into 7 hard screening rules covering: B2B vs B2C mix, recurring vs project split, capex intensity, working-capital profile, customer concentration risk, regulatory/claims risk, and key-person dependence.
List 5 sub-niches to prioritise and 5 to avoid, with one-line underwriting reasons for each.
Provide 10 search keywords + 8 exclusion keywords (Spanish/Portuguese terms welcome).
Output a Lane Scorecard table: Revenue quality / Pricing power / Capex / Working capital / Cyclicality / Operational complexity (score 1–5 + one sentence each).
Output: concise bullets + one table.
2) Associations
Use when: you have a lane and you want a real hunting list, not a thesis.
Prompt
You are helping an ETA searcher build a target universe. Do not invent financials. If size is unknown, label it and propose how to sanity-check it. Important warnings: don’t rely on marketing claims; prefer external directories, regulators, certification bodies, trade associations, buyer “approved vendor” lists, trade fairs/exhibitor lists. Before starting on the task, ask me up to 10 high impact questions that will help sharpen your response.
Lane: [one sentence]
Geography: [countries/regions]
Size band: EBIT €1.5-5.0m (or EBITDA €2-7m).
Preference: privately/family-owned; B2B; low capex; repeat demand.
Tasks:
Produce 12 “where to look” anchors (named categories + examples): associations, certifiers, trade fairs, industry registries, regulator lists, distributor partner lists, franchisor/installer networks, procurement frameworks.
From those anchors, propose 30-50 company candidates with: Company | Country/Region | Why it’s plausibly in-scope | What to verify first (2 checks).
Add a “Sizing sanity check” column: what proxy suggests EBIT €1.5-5.0m (e.g., headcount range, number of sites, contract base, peer margins).
Rules: label each row as Likely / Unclear; no fabricated numbers.
Output: one table.
➡️ Example input (paste as-is)
Prompt
Help me build a target universe for ETA. Do not invent financials. If size is unknown, label it and propose how to sanity-check it.
Rules: prefer external directories (regulators, certification bodies, trade associations, “approved vendor” lists, trade fair exhibitor lists); don’t rely on company marketing. Before starting on the task, ask me up to 10 high impact questions that will help sharpen your response.
Lane: Fire protection inspection + maintenance for commercial buildings.
Geography: Spain + Portugal.
Size band: EBIT €1.5–5.0m (or EBITDA €2–7m).
Preference: privately/family-owned; B2B; low capex; repeat demand.
Tasks:
Produce 12 “where to look” anchors (named categories + specific examples of sources/list types I should search).
Propose 30–50 candidate companies in a table: Company | Region | Why plausibly in-scope | What to verify first (2 checks) | Sizing sanity-check proxy (e.g., headcount, sites, contract base).
Label each row Likely / Unclear.
Output: one table.
3) First owner call prep
Use when: you want the first call to surface fragility, not confirm the teaser.
Prompt
You are a senior strategy consultant advising an ETA investor evaluating an SME acquisition. You will help prepare the first owner call. Important warnings: do not rely on narrative; treat claims as unproven until tied to data; separate Facts / Assumptions / Open Questions; propose cheap verification steps.
Context: [industry], [geography], [revenue/EBITDA range], [business model], [what I know from teaser/website], [ownership/transition situation], [any obvious risks]
Tasks:
List the 8-10 most likely red flags for this deal type (ETA-relevant): revenue quality, concentration, pricing power, key-person risk, working capital/cash conversion, capex/maintenance spend, labour constraints, regulatory/claims exposure.
Provide 12 questions max, grouped: economics; operational reality; customer concentration/pricing; people/capability; succession/motivations.
For each question: what a comforting answer sounds like, what would worry you, and one document/data cut to request next if the answer is weak.Finish with 3 things not to ask yet (and why).
Output: numbered list + short “good vs worry” lines.
➡️ Example input (paste as-is)
Prompt
You are a senior strategy consultant advising an ETA investor. Prepare the first owner call for a potential acquisition.
Rules: treat all claims as hypotheses until tied to data; separate Facts / Assumptions / Open Questions; propose cheap verification steps; don’t invent facts.
Context:
Industry: Fire protection inspection + maintenance (Spain).
Size: €10–20m revenue / €1.5–3.0m EBITDA (range).
Business model: mix of contracted maintenance + callouts + small installs.
Knowns: recurring contracts exist but renewal discipline unclear; technician-heavy; some regulation and liability; top customer unknown; founder still involved day-to-day.
Tasks:
List 8–10 likely red flags for this deal type (revenue quality, concentration, pricing power, key-person risk, WC/cash conversion, maintenance capex/tools/vehicles, labour constraints, regulatory/claims exposure). For each: why it matters and the cheapest falsifier.
Provide 12 questions max, grouped: economics; operational reality; customer concentration/pricing; people/capability; succession/motivations.
For each: comforting answer vs worrying answer + one document/data cut to request next if weak.Finish with 3 things not to ask yet (and why).
Output: numbered list + short “good vs worry + next data” lines.
These prompts don’t make you smarter; they make you harder to fool. We like them because they convert vague “research” into concrete artifacts. If you use them consistently, they push your search into a tighter loop.
Legacy Partners has worked across the ETA ecosystem for years, advising searchers, operators, and investors through acquisition, diligence, and operating transitions. Feel free to reach out by replying to this email! We’re always happy to help.
Insight of the week
A useful underwriting lens for ETA isn’t “how big can this get?” but “how hard is this to kill?” One framework breaks durability into four practical tests:
asymmetry (upside meaningfully larger than downside),
avoidance of single points of failure (“don’t multiply by zero” risks like extreme customer concentration or fragile leverage),
a bias toward protecting the downside over maximizing upside (minimizing unforced errors),
and selecting businesses that are structurally “hard to kill” for a first-time CEO.
In practice, “hard to kill” usually looks like recurring or repeat revenue, low switching costs risk, low key-person dependence, diversified customers, decent EBITDA dollars/margins, and operational simplicity, paired with modest leverage and a thesis that works even if the company mostly keeps doing what it has historically.
Deal watch
Launches
Scrum Capital - ES
Aitor Grandes Gajate and Javier Sánchez Cordero officially launched Scrum Capital. (Link)
Trolega - DE
Tobias Röhrl has officially launched his Search Fund Trolega. (Link)
Transactions
Kaeron - BE
Kaeron (backed by WAD Capital) completed its second HVAC acquisition, adding Chauffage Eric Demory two months after acquiring Groupe Jordan. Demory, founded in 1990, serves high-end residential (B2C) customers across heating/sanitary/water softeners, complementing Jordan’s B2G orientation in the same geography. (Link)
Solferino Capital - IT
Solferino Capital (Francesco Revel-Sillamoni) acquired Tanks International in the late 2025, a distributor of industrial packaging solutions in Italy for chemicals, cosmetics/pharma, industrial, environmental, construction, food/beverage and oil & gas sectors. (Link)
For the commute
Lessons from 22 Off-Market Acquisitions with Peter Lang
Peter Lang completed 22 mostly off-market acquisitions by building a repeatable sourcing and diligence process. In this podcast, he shared how proactive outreach led to better pricing and terms, why he did heavy diligence before LOI to avoid retrades, and how thoughtful deal structures matter more than headline price. The practical takeaway isn’t “negotiate harder”, it’s designing a system that produces more seller conversations, cleaner information earlier, and deal structures that allocate risk where it sits.
Was this email forwarded to you? Subscribe here to get ETA Europe delivered to your inbox every Thursday