Jordan G.

Sharing this in the main channel again to see if others want to join. 🙂 FYI Akhilesh w. is a bot, just ignore. :))

Content should be derived from signals. Playbooks. Sales enablement. Brand. Etc.

We just had a conversation about cross-sell campaigns that relates to this. Our PMMs create these master docs with offshoot resources all about a product – how it fits into the stack, pains and differentiators, competitive positioning, etc. etc. Then we take that information and train agents on them. So idk if I'd agree that we derive playbooks from signals. The "signal", which used to be just "targeting", is essentially the preconditions (which we learn from users and data) for, in the case of cross-sell, someone trying a new product. The content/enablement/etc is a superset, containing the key personas, account profiles, competitive positioning, etc.

I'm sorry but I feel like I'm training your AI and not talking to you 😛

I like it! The signal is tied to a "playbook". That is definitely the next step. But how much you get involved in that depends on your company. I would imagine that the managers would want to own that? Meanwhile, you build the systems, e.g. the automation that lets them know about the risk that they need to execute on.

Akhilesh w. Interesting, that sounds like different levels of pricing risk, definitely a good distinction if your business experiences a lot of low-impact pricing risk (for us, pricing risk is always high impact). Reversibility would get picked up in the model when the risk is resolved; it is still flagged until its resolved (or reversed). Gizem Ç. Haven't we all heard that before 😅 Anything that is flagged in the deal review agent will inform some decision, action, or discussion with the rep or their manager. Anything else can be excluded from the model. If you are flagging real issues, then, provided they trust your insights, if people still aren't taking action then that's a "them problem" lol and that should be on the sales leaders to hash out, but you did your job well.

Gizem Ç. these are the risk categories that the signals roll up into. Then we have a risk category reasoning:

The primary risk is a Pain/Price Mismatch. Company has a clear mandate to move off expensive incumbents like Y and Z, but they are heavily pushing back on pricing. They are seeking a significant cost reduction (aiming for $4M-$8M) which is misaligned with our current proposals based on their usage volumes. This fundamental disagreement on value versus cost is the central point of contention in the negotiation.

Imho, all the signals you are looking to create end up rolling up into a risk category, but open to feedback on that! Akhilesh w. at our company, I just pitch signals, get feedback quickly, and go build / test them. What helps is having a single decision-maker where the buck stops and can just say "we're doing this". Hope that helps!

Evan S. we store raw transcripts in the data warehouse, we also have an embeddings db for use cases that don't require specificity (reduces our costs), and then we store summaries/etc that are generated from transcripts in the data warehouse as well. Hope that helps! Gizem Ç. Good luck with this! Some other thoughts: Anything that applies across multiple deal stages (vague next steps) would be a higher priority for me, and then I'd move on to the signals which apply to the stages where you see the greatest conversion drop off. I am curious how you are evaluating things like vagueness and progress? I'd love to implement this too! I imagine you have some criteria you're looking for (e.g. the next steps did not specify a date or participants, etc.) Yael M. Good call out!

So often, it's not that we need new signals, it's that we need to execute better on what we know works today. That's why I think your focus is super valuable.

Adding to your lists which are great, here's what we do for deal progression: We mine MEDDPICC, was there next steps, new risks identified, do we have EB signoff, etc. from meeting transcripts. Each of these is its own one shot LLM prompt. We then put them in a deal risk dashboard so managers can see which deals are on track. Recently I started summarizing all meetings, emails/calls, and slack msgs to create a bullet point executive summary. For future context purposes, I also extract whether there were next steps. These can be used for future alerts, or as context in conversational agents (e.g. automatically runs workflows to complete each next step). I shared the full prompt on substack if you'd like! Planning on writing about the deal progression system soon too.

Good idea!