AI Xccelerate sells proof before commitment, yet its public proof has no numbers. All four case studies are anonymized and none gives a result. The one live deployment, support triage for a healthcare software company, marks its monthly ticket volume and deflection rate as "reserved". The proof page defines four evidence levels (Demonstrable, Deliverable, Operational, Business) but does not say which level each story has reached. The firm's own comparison page notes that a smaller rival names client results while AI Xccelerate shows four anonymized stories, and enterprise rivals such as Fractional AI publish named results like an 80% cut in hallucinations.
First action
Email the support-triage client this week asking to approve publishing an anonymized deflection rate, with a draft card attached.
How to measure it
Proof-page form submissions
- Now
- –
- Target
- 2x the current monthly count
- By
- 90 days
Steps
Ask for approval
Send the support-triage client a one-page request: what will be published, anonymized, with their sign-off on the wording.
Founder & CEO 2 days
Fix the baseline
Pull pre-go-live weekly ticket volume by channel and the share answered from the knowledge base by people.
Forward-deployed engineer 3 days
Measure
Four weeks of tickets closed by the AI worker without a person, tickets routed, and reopen rate as a quality check.
Forward-deployed engineer 4 weeks
Write the result card
Number, period, method, attribution, limits and evidence level, in the format the proof page already promises.
Marketing lead 2 days
Label all four stories
Add a Demonstrable, Deliverable, Operational or Business badge to each story and to the case-studies index.
Web developer 1 day
Place it
Result card on the homepage under the hero, at the top of /case-studies and /proof, and on the comparison pages.
Web developer 1 day
Use it in sales
Add the card and the evidence-review offer to every first-call follow-up.
Founder & CEO ongoing
Repeat
Ask each of the other three clients for one approved number in the next quarter.
Delivery lead 1 quarter
Why it works
Use the method the proof page already describes (baseline, method, attribution, limitations, permission) and apply it once, to the support-triage story. Agree a baseline with the client, such as weekly tickets and the share closed without a person before go-live, and measure four weeks after go-live. Publish one result card with the number, the period, how it was measured and its limits. The client can stay anonymous. Then tag all four stories with their evidence level, so a buyer sees at a glance which are Demonstrable, Operational or Business. Put the result card at the top of /case-studies, /proof and the homepage, and offer the full evaluation pack through the existing "Request an evidence review" button.
Enterprise buyers shortlist on evidence. One honest, well-documented number from a live system is more credible than four stories with none, and labelling each story by evidence level turns AI Xccelerate's rigor into a visible advantage instead of a reason the stories look thin. The anonymized format means the client only has to approve the number, not its name.
Ready to use
CopyVerified result card
STORY: Support intake that deflects and routes (healthcare software, EAP services) EVIDENCE LEVEL: Operational -> Business once approved RESULT: __% of inbound tickets closed by the AI worker without a person PERIOD: [4 weeks, dates] vs baseline [4 weeks before go-live] CHANNELS: phone, chat, message, email METHOD: tickets closed with no human touch / all tickets; reopened tickets counted as not closed ATTRIBUTION: no other support changes during the period LIMITS: one client, one knowledge base, early weeks APPROVED BY: client (anonymized) on [date] CTA: Request an evidence review
Risks and how to handle them
The client declines to publish
Offer a range (for example "30-40% of tickets") or share the number only through the evidence review under agreement.
The early number is modest
Publish it with its trend and limits; an honest first figure matches the gate philosophy and can be updated.
Evidence
- Your websiteSupport-triage story: an AI worker takes intake across phone, chat, message and email and closes low-complexity tickets from the client's own knowledge base; the deployment is "in flight"; monthly ticket volume and deflection rate are marked "reserved" until they can be verified; the only stated result is "Support capacity is recovered without adding people."Source
- Your websiteProof page: four evidence levels (Demonstrable, Deliverable, Operational, Business); named or verified metrics appear only after approval of baseline, method, attribution, limitations and publication permission; no outcome figures are published yet; sanitized evaluation results and references are offered under agreement through a "Request an evidence review" button.Source
- Your websiteCase studies page: "Four problems. Four workers. Four places a prospect can land." All four stories are anonymized (an Australian not-for-profit serving elderly clients, an AI software company in the Microsoft partner ecosystem, a healthcare software company running EAP services, a Mexico-based promotional products company entering Spain and Italy) and none publishes a result number.Source
- Your websiteComparison page vs PxlPeak: PxlPeak publishes voice plans from $297 a month and managed plans at $997, $2,497 and $4,997 a month; the table says AI Xccelerate does not publish prices, and that PxlPeak names client results for its web and ecommerce work while AI Xccelerate shows four anonymized stories.Source
- Competitor researchFractional AI's homepage names customers such as Zapier, Airbyte and LiveRamp and publishes results such as "Reducing Hallucinations by Over 80%" (Zapier) and "10x'ing the Speed of Building Connectors" (Airbyte); it has separate pages for enterprises and private equity firms.Source