AI training for private equity:
Ciclad's Invest team moves from prompts to skills
On 27 March 2026, Emma Pariente kicked off Ciclad's AI programme with a generative AI Show: ChatGPT, Claude, Perplexity, the art of prompting and confidentiality, all applied to the day-to-day work of an LBO fund.
The Invest team then came back for a role-based workshop led by Loïc Dworzak. On the agenda: MCP connectors, Claude Skills and Projects, and 6 deal scenarios to solve with AI.
The goal: give every task that comes up on every deal (NDAs, LOIs, tax returns, data rooms, LBO models) its own tool, reusable by the whole team.


The client
Ciclad is an LBO fund specialising in French SMEs, with expertise in buyout and succession capital. Its Invest team works on long, sensitive documents: information memoranda, tax returns, data rooms and due diligence reports.
The challenge
From one-off prompts to tools shared by the deal team
In private equity, the same tasks come up on every deal. An NDA to review that same evening. Three years of tax returns to restate. An LOI to send before 6pm. A data room of several hundred documents to comb through before the committee.
The challenge set for AI Sisters: first bring the whole team up to speed on generative AI, then get it building reusable tools for its own processes. Without ever losing control of the numbers.
Long, varied documents
100-page IMs, 150-page DD reports, tax returns in PDF, messy Excel files.
Tight deadlines
A committee on Thursday, an LOI tonight, an LBO model for 11am tomorrow. Analysis time is always short.
Confidential data
Every deal involves sensitive information about a target. Choosing the tool and its level of confidentiality is not a detail.
A demand for reliability
In M&A, a hallucination is costly. What the AI produces must be sourced and open to challenge.
Is your deal team still reviewing every NDA by hand?
We build with them the skills that handle the recurring work, on your own templates.
Our approach
Our AI training programme for Ciclad's Invest team
Two stages: a Show to build AI awareness, a role-based workshop to build tools. Every example and every exercise is drawn from the day-to-day work of an LBO fund.
The generative AI Show
Emma Pariente lays the foundations: what an LLM is, the vocabulary of AI, what ChatGPT can do today. Then the structure of an effective prompt (role, context, task, goal, constraints and format, tone), refining answers and reverse prompting. A questionnaire sent beforehand helped tailor the content: 92.3% of the 15 respondents already saw AI as essential in their sector.
Prompts at the ready, on Ciclad cases
No generic examples. Participants draft a price renegotiation email after a due diligence that uncovers an unprovisioned employment dispute. They run a market benchmark with Perplexity. They ask Claude for an interactive artefact to understand a sector they are discovering in a teaser.
Confidentiality, level by level
Five levels of data protection, from the lightly regulated free tool to local AI and dedicated servers, by way of enterprise versions. Then the risks specific to the profession: hallucinations, bias, content ownership. The rule adopted: always cross-check against reliable sources.
The Invest team role-based workshop
Led by Loïc Dworzak, the workshop moves from prompt to tool. MCP connectors to plug Claude into the team's tools. Skills for reusable instructions. Projects to share documents and context. Then 6 scenarios modelled on real deal situations: the NDA due back that evening, 180 pages of tax returns for an LBO model the next day, an LOI going out at 6pm, 350 documents in a data room before the committee. Each time, the same mission: deliver a Claude skill or project, tested on the case, usable by the whole team. And the same instruction: never blindly accept what the AI generates, but challenge its calculations.
A Show to bring the whole fund on board, a workshop to take action?
We adapt both formats to your investment processes.

Use cases
AI use cases in an LBO fund
More than twenty use cases reviewed with the Invest team, from the first teaser to the investment committee.

Screening and sourcing
→ Pre-scoring an incoming deal against Ciclad's criteria: pursue or pass in 5 minutes
→ Express screening of a 2-page teaser
→ Deep Research to map out a niche market in 20 minutes
→ Monitoring and qualifying M&A intermediaries
Deal analysis
→ Structured summary of an IM, then Claude as devil's advocate to audit it
→ Accounting extraction from a small-cap patchwork, with a source column for every figure
→ From a PDF tax return to a pre-filled Ciclad Excel file
→ LBO model pre-filled from the IM
Due diligence and data room
→ Sourced red flags and Q&A on a data room
→ Summary of a financial DD report: traced EBITDA restatements, price adjustments, discrepancies with the IM
→ Mapping employment risks and the commercial side before the external DD
→ Updating the DD Q&A tracker with each incoming answer
Legal documents and committee
→ NDA reference grid built from 16 past NDAs, then a Ciclad NDA pre-filled in 2 minutes
→ Descriptive sections of an LOI, an IC memo, a two-pager or a binding offer
→ Investment committee minutes ready 15 minutes after the meeting
→ 6-slide pre-committee deck generated from the IM

Results
The results of the AI training at Ciclad
complementary formats: an AI upskilling Show and a role-based workshop
private equity use cases reviewed in the workshop
IMs, data rooms, LBO models: where to start?
Let's talk about your deals and your confidentiality constraints.

















.jpeg)






















































































.png)
.jpeg)



























.avif)