ACTUALIZATION.AI
WORK, FUNDING, AND RESEARCH WITH
WHAT
Short engagements: diagnosis, advice, and training.
FOR
Teams shipping LLM, RAG, NLP, or agent features.
OUTPUT
Evidence, root causes, and prioritized fixes your team can act on.
Most AI budgets are spent on the wrong thing.
95%
of enterprise GenAI pilots produced no measurable impact on profit or loss, against $30–40 billion of spending.
MIT looked at 300 public deployments, 52 organizations, and 153 executives, and found the gap was not model quality or regulation. It was approach. Only about 5% of pilots reached production with value anyone could measure. The mistakes behind the other 95% are specific, repeated, and avoidable.
WHERE THE MONEY GOES WRONG
The project starts because AI is expected, not because a business problem is defined.
Budget goes to sales and marketing demos while the measurable returns sit in operations and finance.
A generic tool performs in the demo and breaks in the real workflow.
Nothing was baselined, so nobody can prove whether quality moved.
The work is built in-house from scratch; MIT found vendor-built tools succeeded about twice as often.
Scaling drags on for nine months, the average for large firms, against ninety days in the mid-market.
Start with the smallest engagement that answers the question.
Scope, timeline, and deliverables are fixed in writing before work starts.
START SMALL
One-Hour Consultation
$149
1 hour
A working call on your system: what it is doing, what is most likely causing it, and whether a larger engagement is worth your money.
Not sure yet? Book a free 15-minute chat first.
Book an Hour
DIAGNOSE
AI Behavioral Diagnosis
FROM$5,000
about 5 business days
Reproduce the issue, test the likely causes, and deliver a root-cause analysis with prioritized fixes.
Start a Diagnosis
MEASURE
LLM / RAG Evaluation Audit
FROM$10,000
1 to 2 weeks
Build the evaluation set, establish a baseline, classify failure modes, and produce a measurable improvement plan.
Request an Audit
TEACH
Talk or Short Course
Contact me
one session or a series
A talk or short course for your organization on how these systems behave, where they fail, and how to evaluate them.
Discuss a Talk
You receive evidence, not a strategy deck.
Findings come back as a written root-cause report and a live readout with your engineers. Every row is reproducible, ranked by severity, and paired with the first repair I would make.
Your team can act on it without me. That is the point.
AREA
FINDING
SEVERITY
FIRST REPAIR
{{ f.area }}
{{ f.finding }}
{{ f.severity }}
{{ f.repair }}
GOOD FIT
✓You already have an AI system, prototype, or vendor implementation.
✓You can provide representative inputs, outputs, logs, or system access.
✓A measurable quality, reliability, cost, or behavior problem exists.
✓You want evidence and a prioritized plan, whether your team executes it or I help.
NOT THE RIGHT FIT
·You want a generic AI strategy presentation.
·You have no defined workflow, user, or business problem.
·You need an enterprise transformation or a staff augmentation program.
·You expect a guarantee that a probabilistic system will never be wrong.
The person who will be on your engagement.
{{ p.name }}
{{ p.role }}
{{ p.bio }}
Based in Tampa, FL, USA. No account managers, no offshore bench. You work directly with the person doing the work.
Proof, in order of what usually matters most.
Most AI failures are reasoning failures wearing a software costume. Telling the difference takes someone who has studied reasoning for twenty years and also had to make a shipping product behave.
01
A live AI product
SquarePact is my agentic document-intelligence platform for Microsoft Word workflows. I operate it, support it, and feel every reliability problem in it before a client ever does.
02
Public-sector delivery
A $100,000 pilot and co-development partnership with Tampa International Airport / HCAA. Final wording and logo use pending client approval.
03
Enterprise consulting
Prior consulting work for The Appraisal Foundation. Published here only in language the client has approved.
04
Funded research
NSF SBIR-funded work on automated and correct reasoning over rules and legal text, plus research at the USF Advancing Machine and Human Reasoning Lab.
05
Research record
More than 100 peer-reviewed publications across 20 years, concentrated on reasoning, argumentation, and language understanding rather than general commentary about AI.
I describe my university relationship accurately. USF does not endorse this consulting practice, and I do not imply that it does.
How I handle your system and your data.
NDA first
I sign an NDA, subject to review, before you send anything sensitive. Nothing confidential should ever go through a public web form.
Least access that works
Many diagnoses run on API access, logs, traces, and test cases. I ask for source or production access only when the finding depends on it.
You own the output
Reports, test sets, and any code I write during an engagement are yours. Your team can act on the findings without me.
PROOF THAT I SHIP
From the people behind
SquarePact is my agentic AI platform for complex document analysis and Microsoft Word workflows. I do not only advise on AI systems. I design, build, deploy, and operate one, and I carry that experience into every diagnosis.
Explore SquarePact ↗SERVICE · AI BEHAVIORAL DIAGNOSIS
Find out why your AI is not behaving as expected.
I reproduce the problem, test the likely causes, and deliver a diagnosis with prioritized corrective actions. A professor of AI does the analysis, informed by running a production AI product day to day.
PRICEFrom $5,000
DURATIONAbout 5 days
SCOPEOne existing LLM, RAG, NLP, or agent workflow
STARTS AFTERAccess, examples, and a signed scope
Book a Free 15-Minute Consult
See what I examine
What I examine
{{ item }}
What you receive
{{ item }}
NOT INCLUDED
A guarantee that the system can be made perfectly accurate.
Implementation of the recommended repairs.
A full security, privacy, regulatory, or penetration audit.
A complete production evaluation framework.
New application features or user-interface work.
Large-scale data labeling or dataset creation.
Unlimited meetings, revisions, or added workflows.
Five steps, about five business days.
01
Intake
You provide the expected behavior, the observed failures, representative examples, and available access.
02
Reproduction
I confirm the problem and define a small set of cases that captures it.
03
Diagnostic experiments
I test model, prompt, data, retrieval, tool, orchestration, and architecture hypotheses against each other.
04
Root-cause analysis
I identify the most likely causes and separate symptoms from underlying failures.
05
Findings readout
You receive the evidence, the recommended repairs, and a live readout to walk through them.
A findings excerpt, in the format you get.
AREA
FINDING
SEVERITY
FIRST REPAIR
{{ f.area }}
{{ f.finding }}
{{ f.severity }}
{{ f.repair }}
Questions I get before signing.
{{ q.q }}
{{ q.a }}
Tell me what you're working on. I'll tell you if I can fix it.
Free 15 minute consultation to see if we're a fit.
✓A senior technical read on the behavior you are seeing.
✓I take a limited number of engagements, so I will say if this is not one for me.
✓Scope and price are fixed in writing before any paid work starts.
REQUEST RECEIVED
Thanks. I respond within one business day.
You will get an email confirming what you sent, along with a link to pick a fifteen minute slot. If the behavior you described is not something I can help with, I will tell you that in the reply rather than on a call.
Something urgent? Write to john@actualization.ai.
SERVICES
One-Hour Consultation
AI Behavioral Diagnosis
LLM / RAG Evaluation Audit
Talk or Short Course