PXN Group: an AI first pass on 480 board packs a month

PXN Group receives four packs a month from each of its 120 portfolio companies, plus hundreds of pitch decks a year from prospective investments. OBT built an application on Azure OpenAI that reads them all, extracts the highlights and gives every board rep a concise, consistent briefing before the meeting.

In their words

We have 120 portfolio companies that send us four packs every month. Obviously that is a huge number to have to review and go through. So what we were after from OBT was a system that helps us do a first pass of them, so all of our board reps get to board meetings fully prepped.

The system should be able to take all packs in from emails and do a first pass by reading all the information, pulling out some key highlights and context, but also completing that review in line with a look back over previous periods to ensure there is a level of consistency.

We can then present the board member, ahead of the meeting, with a concise summary and some points to pick up on, including some questions that we feel need answering. When you add that to the information which the board member already knows, hopefully this means that we can go in fully prepared, having saved time on a previously manual board pack process. It also means that over time we have a bank of data that is uniformly stored, an AI that is getting smarter, and we can use that to add more context to what we do in board meetings.

A secondary challenge that we needed support with is the hundreds, if not thousands, of slide decks we receive a year for potential investments.

AI does a quick scan of all of these to understand the detail of what is in there, what the business does, extract some key highlights and use this information to support the ventures team in doing the first pass of these pitch decks. We can then decide whether the business opportunities are ones that we would like to progress with, or we need to find out more information. Traditionally this is quite a labour intensive job, so we are using AI to cut that down.

Thomas Hardman, Insights Director at PXN Group

What we did

To help with the ever increasing number of clients under PXN's portfolio, OBT built an application that harnesses the power of OpenAI on Azure to extract and enrich pitch decks and board packs from prospects and portfolio companies respectively.

This removes the manual effort of trawling through 120 plus variations of complicated tables, board pack material and targets, to give PXN a succinct, thorough report on the state of a business or the likelihood of success for a pitch deck.

The application covers two workflows:

• Board packs. Packs arrive by email and are read automatically. The system pulls out key highlights and context, compares them against previous periods for consistency, and produces a summary plus suggested questions for the board rep ahead of the meeting.
• Pitch decks. Incoming decks are scanned to understand what the business does and to extract the key highlights, supporting the ventures team through the first pass so they can decide quickly what to progress and where to ask for more.

Over time, both workflows build a uniformly stored bank of data that adds further context to every board meeting and investment decision that follows.

---

Written by Olivia Jaskolka · Edited & proofread with AI assistance