fieldnotes
ACE deal sizing from estimates (with and without AI)
2026-01-16
What’s this one all about then ?
Back in action in the New Year, and trying to work out what we missed between pre:Invent, re:Invent, end of year, Christmas/New Year holidays, and we spotted this this post-re:Invent announcement about being able to generate or compare deal sizing estimates with AI in PartnerCentral in the ACE workflow.
This new feature, available within the APN Customer Engagements (ACE) Opportunities, uses AI to provide deal
size estimates and AWS service recommendations. Deal Sizing capability allows Partners to save time on deal
management by simplifying the process of estimating AWS monthly recurring revenue (MMR) when creating or updating
opportunities.
Partners can optionally import AWS Pricing Calculator URLs to automatically populate AWS service selections
and corresponding spend estimates into their opportunities, reducing the need for manual re-entry. When a Pricing
Calculator URL is provided, deal sizing delivers enhanced insights including pricing strategy optimization
recommendations, potential cost savings analysis, Migration Acceleration Program (MAP) eligibility indicators,
and modernization pathway analysis. These enhanced insights help Partners refine their technical approach and
strengthen funding applications, accelerating the funding approval process.
So of course, we took a look.
AI-suggested MRR
First of all, we just looked at the AI-suggested MRR based on the opportunity.
Scenario 1: We edited an existing ACE opportunity which had an arbitrary MRR (it was a professional services opportunity, where MRR is extremely rubbery) and no service selections. Our suggested MRR for this opportunity was $480.
Somehow, AI decided that $372 was more appropriate. Given there was no service selections and it was pro serv, this seems pretty random.
So we tried it with a new ACE opportunity with the same title/description/customer/solution and arbitrary MRR $500. Somehow, AI decided this one should have been $480.
Wierd.
Scenario 2: Again we edited an existing ACE opportunity with an educated MRR (it was a bundled proserv + infra opportunity, with service selections). There were similar ACE opportunities already closed lost and launched, all linked to the same solution.
The MRR was $750. AI decided that $926 was more appropriate.
Comparatively, the average MRR of closed lost and launched opportunities for scenario 2 was $725. So somehow, AI decided that this one should be bigger. The only factor we can put it down to is perhaps something from the customer’s profile which influences the upgrade.
Scenario 3: We created a “moderately well described” ACE opportunity with a $1 MRR. This time AI decided that $230 was more appropriate. That’s quite an upsell …
Scenario 4: We created a “moderately well described” ACE opportunity for software sale via AWSMP, and gave it an MRR based on 100 users x the vendor’s public AWSMP pricing of $35 per user per month - so MRR $3,500.
AI decided that it should be $22,800. Yes, thats right, about 8x larger. We think it picked the enterprise license which is advertised on AWSMP public pricing at $25,000 as the base and somehow discounted it.
Our take aways from this:
- The AI-suggested MRR ranges from reasonable to something which is very unreasonable.
- It appears that the AI MRR suggesting mechanism is influenced by a range of factors, including the partner’s similar opportunities, the customer’s opportunity history, scale indicators in the description of the opportunity, the service selection mix, and AWSMP public pricing.
- Compare your own hand-developed MRRs against AI, but don’t ever take the lazy default of accepting its recommendation. If the MRR on an ACE opportunity is ‘maleable’, at least you know what went into that, with AI, you have no idea of the provenance.
Importing Pricing Calculator URLs
Next, we used the pricing calculator import.
Scenario 1: a moderately well specified calculator against a well defined solution with linked services - completing most options to completely define a classic 3 tiered web data ingest and analyse application (Amplify, RDS, S3, Glue, Athena, EC2 Fargate, Lambda, Quicksight). The pricing calculator came in at MRR $183, and we imported it into the ACE opportunity.
AI-adjusted it was recommended to be MRR $192. That’s pretty close although given that it was well specified, its not clear where the 2-3% bump came from.
Scenario 2: a sparsely specified calculator vs a well defined solution with linked services - a bare bones S3, EC2, RDS. The pricing calculator came in at MRR $94, and we imported it into the ACE opportunity.
AI-adjusted it was recommended to be MRR $53. That wasn’t what we expected, we thought AI would bump the number up, realising that the solution specification was larger - not least we had selected a few services for the opportunity that were not represented in the pricing calculator.
Our take aways from this:
- When a pricing calculator is used, the AI suggested MRR is “adjusted” and not always in the way we would expect.
- Even when a well specified pricing calculator is used, for some reason the AI feels compelled to make an adjustment
Mysteries
There are still more than a few mysteries:
- What kinds of input and algorithm are used by AI to estimate/suggest the MRR ? We’ll probably never know, but what we can say is that certain things seem to have impacts, and also that AI can be a bit mislead in both the historical ACE opportunity case and the pricing estimator case
- How is partner provided MRR going to sit alongside AI-suggested MRR in internal AWS seller systems ? Its a logical thing to do, and AWS internal systems already have various weightings and scorecards applied to partner opportunities to support resource allocation and stack ranking.
Learnings, better practices reinforced
There are some important reminders from the availability of this new feature:
- Reinforces the practice of ACE opportunities being (extensively) tagged with service selections, which potentially now inform AI suggested MRR calculations.
- Reminds us that pricing calculators should be used wherever possible in association with an opportunity. Its easy to slip out of that habit, because adding in a random MRR is easier than using the price calculator, especially early on.
The wrap up
A lot of vendors have introduced AI opportunity sizing support, but in our experience few with as diverse a portfolio as AWS, let alone a portfolio that mixes products and services, and a marketplace. Its quite a challenge, but on the other hand it does align with the emerging AWS perspective if not priority to be more focused and disciplined about MRR and the commercial impact that partners have overall.