# AI Ops Diagnostic: Find the AI Work Worth Doing | Super Cat

> Super Cat Technology reviews one business process, checks the data, and ranks the AI and non-AI projects worth doing. Vendor neutral, run by builders.

Source: https://www.super-cat.tech/ai-ops-diagnostic  
Last modified: 2026-10-22

AI Ops Diagnostic

# Find the AI projects worth doing before you build anything

Written by [Cat Yung](https://www.super-cat.tech/cat-yung), Founder, Super Cat Technology. Last updated 22 October 2026.
Before anyone builds, Super Cat Technology reviews how one of your processes runs. We map where time and money leak, check the data the process depends on, and rank the projects worth doing, AI or not.
The output is a ranked shortlist with the reasoning behind it. It does not commit you to a build, and some findings may point to fixes that need no AI at all.
Book a First-Sprint Data Scientist [See What We Have Shipped](https://www.super-cat.tech/portfolio)

## Built for the person who owns the operational result

You own collections, intake, customer calls, document handling or reporting. You know the work is slow or expensive. You have seen the AI demos. You don't yet know which project would pay back first, or whether your data can carry it.
Not for teams that already have a scoped build and need engineers now. For that, see [hire vetted AI engineers](https://www.super-cat.tech/hire-vetted-ai-engineers).

## A vendor-neutral AI implementation consultant that also builds

Vendor neutral means we recommend what fits your process, including tools we don't build or sell. We look at the work first and the technology second. If the best fix is a plain software change, the shortlist says so. Builders run the review, so every option on it is one we know how to ship.
Weighing a part-time AI lead instead? Our guide to [fractional CTO vs CAIO vs full time](https://www.super-cat.tech/guides/fractional-cto-vs-caio-vs-full-time) sets out that choice. Our [fractional CTO roadmap review](https://www.super-cat.tech/portfolio/fractional-cto-review) is the closest published example of this kind of work.

## What you provide, and what we do

| Super Cat does | You provide |
| --- | --- |
| Runs a structured intake call on the process and its numbers | One owner who can answer questions and decide |
| Walks through the process with the people who do it | Access to two or three of those people |
| Checks a real sample of the data the process depends on | A representative sample, not a cleaned extract |
| Maps where time and money leak, step by step | Any cost or volume figures you already track |
| Ranks the candidate projects, AI and non-AI, with the reasoning | Your view on constraints: budget, policy, systems |
| Presents the shortlist and answers questions | Thirty minutes with the decision maker |

## Review first, build later

The process decides which project is worth building, so the review comes first. There are four stages.
Stage 1

### Intake

We agree the process, the owner and the questions.

Stage 2

### Walk-through

We sit with the people doing the work.

Stage 3

### Data check

We test a real sample against what each project would need.

Stage 4

### Readout

We present the ranked shortlist.

Access delays and a missing owner are what usually stretch it. You don't need to prepare a brief or choose a vendor first.

## The team that reviews the process has shipped what comes out of it

An [AI payment collection system](https://www.super-cat.tech/portfolio/ai-payment-collection) we built recovered US$1M in its first 3 months. A voice agent we built handles 200k calls a month for a 5,000-employee US enterprise. Our fractional CTO roadmap review was rated 5.0 by the client.
Cat Yung is Expert-Vetted on Upwork, top 1% platform-wide, with a 100% Job Success Score (source: Upwork profile, October 2026). She has over a decade in production AI, NLP and ML. See [all case studies](https://www.super-cat.tech/portfolio) or [Cat's profile](https://www.super-cat.tech/cat-yung).

## Where AI projects go wrong

> “They would like to use AI to optimise their workflow and earn more money. They should be building their own internal knowledge system and skills, to understand their pain points and priorities, but they focus on the AI framework instead of the internal logic.”
A diagnostic starts with the documents, rules and know-how the business runs on, and with the pain points of the people doing the work. The framework and the model come later.

## An example: the AI call centre

A 5,000-employee US enterprise needed its phones answered around the clock across 60 HVAC brands. We built AI voice agents that book appointments and hand off to a person when confidence drops. They now handle 200,000 calls a month, booking 67% of calls against 77% for the human team. Read the [full case study](https://www.super-cat.tech/portfolio/ai-agent-call-center).

## Vendors changed the ground under AI projects this year

Models are retired on fixed dates, and data terms change. OpenAI set 23 October 2026 as the API shutdown date for GPT-4, GPT-4 Turbo, o1, o3-mini and o4-mini, after notice in April. Anthropic gave two months' notice that Claude Sonnet 4.5 retires on 30 November 2026. From 17 August 2026, Atlassian began training its AI on Jira and Confluence content by default for Free and Standard plans.
The lesson for a first AI project is to plan for the model to change. The prompts, tests and business rules should belong to your team, so a model swap is a planned task and not an outage.
Sources: [OpenAI API deprecations](https://platform.openai.com/docs/deprecations) (notice of 22 April 2026) · [Anthropic model deprecations](https://platform.claude.com/docs/en/about-claude/model-deprecations) (notice of 30 September 2026) · [Seibert Group on Atlassian's AI training policy](https://us.seibert.group/blog/atlassian-data-privacy-ai-training-policy-change) (23 April 2026).

## Risks, and who handles each

### The data is not what it looked like.

A sample can show the records are thinner than the first call suggested. We check a real sample early and tell you what it supports. You share a representative one.

### No owner on your side.

We raise it at intake and pause until one person can decide. You name that person.

### The best answer is not AI.

We rank non-AI fixes alongside AI ones. You treat that as a result.

### Sensitive data needs a different home.

We flag private or regulated documents at intake and scope around your policy. Your compliance owner confirms the rules.

Accountable by name: your diagnostic is led by Cat Yung, Founder of Super Cat Technology, who signs off the shortlist.

## When the shortlist touches private finance documents

Some projects touch documents a family office or finance team keeps in-house. For those, [Super Chain](https://www.super-chain.tech/) is our AI platform for family offices. It answers questions over the office's own documents and cites the source for each answer.

## Questions buyers ask

### What does a vendor-neutral AI implementation consultant do?

A vendor-neutral AI implementation consultant recommends the tool or approach that fits your process, not the one they resell. They review the work first, test your data, and say plainly when a cheaper or non-AI option wins. Ask them which tools they have shipped and which they would refuse to recommend.

### Do I need an AI consultant who actually builds, not just strategy?

Yes, if you want the plan to survive contact with your data. A consultant who builds knows what a project costs to ship and run, so their ranking reflects real effort. Super Cat’s reviewers have shipped payment collection, voice agents and document search in production, so each option on the shortlist is buildable.

### Can an AI consultant rescue a failed AI project?

Often, if the failure came from the wrong project or thin data rather than the model. The first step is the same as a diagnostic: check what the process needs and what the data supports. Sometimes the answer is to narrow the scope. Sometimes it is to stop and pick a different project.

### Hands-on AI consultant vs consulting firm: which one do I need?

A consulting firm suits a company-wide programme with many workstreams and a large budget. A hands-on AI consultant suits one operational problem that needs a clear answer and then a build. If you can name the process that hurts, start hands-on and bring in a firm later if scope grows.

### How do I evaluate an AI consultant before hiring?

Ask for shipped work you can check, not slides. Look for production results with dates, client ratings on a platform that verifies them, and a willingness to say a project is not worth doing. Ask who will do the work, and what they need from your team in the first week.

### What do I need ready before an AI operations review?

Have one owner who can decide, a rough sense of volumes and costs for the process, and access to two or three people who do the work. A real data sample helps most. You don’t need a brief, a budget line for AI or a preferred vendor.

## Start with the process that costs too much.

If you own a process that costs too much and you're not sure where AI fits, book a call. On the first call we listen to how the work runs today. Then we tell you whether a diagnostic is the right first step, before any fee comes up.
Book a First-Sprint Data Scientist

First-Sprint Data Scientist

## A vetted engineer embedded in your team, a shipped deliverable by end of sprint 1.

Pick a time below and you will speak with me directly. Bring the use case, the data and the constraints, and I will give you a straight view on feasibility, effort and the first thing to ship.
Cat Yung
Hong Kong & London time zones

[Open the scheduler](https://calendly.com/super-chain/catyung?utm_source=super-cat.tech&utm_campaign=first-sprint-data-scientist&utm_content=ai-ops-diagnostic)