Leadership perspective

Why our website tells you what is not ready yet

Why UniFlow labels what is sample, partial or not ready, and a calm guide to evaluating AI software claims before you trust a demo or a product page.

Zeeshan KhanCMO, UniFlow5 min read

A buyer sits down to compare a handful of workspace products. Every product page reads as if the software is finished. Every page mentions AI. Every demo video runs smoothly. By the end of the afternoon, the buyer has learnt a great deal about how each company wants to be seen, and very little about what each product does today. Evaluating AI software claims has become hard work, and buyers are right to be cautious.

This piece reflects the market perspective UniFlow’s CMO, Zeeshan Khan, leads on for the team. It explains why our website labels what is not ready, and how we think buyers should read software claims in general.

The problem with pages that sound finished

Marketing pages are written to persuade, and there is nothing wrong with that. The difficulty comes when persuasion blurs status. A feature that works only with sample data, a screen that shows a future direction, and a capability that is fully built can all be described in the same confident sentence. The buyer has no way to tell them apart until after purchase.

AI makes this harder. An AI feature can look convincing in a demo while running on a sample provider, on hand-picked examples, or with a person quietly correcting the output. None of that is dishonest in itself. It becomes a problem only when the buyer is not told.

The cost is not limited to one disappointed customer. Each gap between what a page promised and what a team found erodes trust in the whole category, and makes the next buyer more sceptical of every vendor.

Why we label status instead

At UniFlow, we decided that clear status labels are more useful to a buyer than polish. We would rather a team read our pages, understand exactly what is ready, and then decide, than discover the gaps after a commitment.

So we keep a plain disclosure next to our claims. UniFlow is available for guided evaluation. Some AI and advanced workspace screens use sample behaviour. External AI providers need configuration. A demo is not proof of production readiness.

We also describe each capability by its actual state. Here is a short version:

Status Examples in UniFlow
Implemented Project channels and threaded messages, accepting or rejecting proposed decisions, keeping ideas and promoting them into tasks, task owner, status and priority, goals and milestones, lead stages, in-app notification inbox
Partial Conversation processing that proposes decisions and ideas (sample AI provider by default), connected search and ask, meeting records without verified live video or audio, workspace member roles, mobile direction
Concept or sample Visual knowledge map with hand-authored sample nodes, brainstorm and debate screens, automation administration screens, a revenue screen using sample data
Not ready Phone push delivery

The full breakdown sits on our trust page.

Evidence boundaries matter as much as features

A feature list tells you what exists. Evidence tells you how far you can rely on it. These are different questions, and buyers should ask both.

For UniFlow, the evidence we can point to is a local twelve-stage core-loop exercise. It covered messages, extraction-shaped output, stored insights, relationships, retrieval, notifications and cockpit results. It used a deterministic sample AI provider. That validates a local path through the product. It does not prove real-model quality, and it does not prove performance at production scale. We say so, because a buyer who knows the limits of the evidence can plan around them.

We do not show customer logos, testimonials, certifications or performance figures, because we have none to show. We do not publish a price; we provide a written scope and price on request. Saying less than a buyer might expect is better than saying something we cannot support.

A short guide to evaluating AI software claims

Whichever vendor you are looking at, these questions help separate what is shown from what is ready:

  1. Which AI provider ran this demo? Was it a sample provider, a configured model, or a person correcting output behind the scenes?
  2. Who reviews the AI’s output before it counts? If the answer is nobody, ask how errors are caught.
  3. What is the status of each screen? Ask for labels: implemented, partial, sample, concept or not ready.
  4. What evidence supports the claim? A local test, a pilot, a production deployment and an independent audit are very different things.
  5. What is not claimed? Notice what a vendor avoids saying. Silence about customers, certifications or scale is information too.
  6. What will it cost, in writing? A verbal number is not a price.

Our guide on how to evaluate a team workspace turns these into a practical checklist for a guided session.

Why human review is part of our design

Labelling status on the website is one half of our approach. The other half is in the product itself. In UniFlow, AI can propose decisions and ideas from a conversation, but a person must accept or reject each proposal before it becomes a record the team relies on. That is a deliberate choice: an explicit review step rather than automatic commitment. We explain the reasoning in review before you commit.

The two halves come from the same belief. AI output should be treated as a proposal, and a buyer’s trust should be earned one verified claim at a time.

What this means for buyers

We are aware that this approach can make our pages look less impressive than some. A reader scanning quickly may see the words sample, partial and not ready and move on. We accept that. The teams we want to work with are the ones who read carefully, test with their own examples, and decide on evidence. For them, a clear map of what is ready saves time and avoids an unpleasant surprise.

If you want to see how this works in practice, the evaluation page sets out how a guided session with UniFlow runs, and how it works explains the four stages of the workflow.

Takeaway

Honest labels are not a weakness in software marketing. They are what lets a buyer trust the claims that remain. When you read any product page, look for status, ask which AI provider ran the demo, check who reviews the output, and ask what evidence stands behind each claim. At UniFlow, we label what is sample, partial and not ready so that you can make that judgement with the facts in front of you.

Written as a UniFlow company perspective. No direct quotations are attributed. Byline subject to the named author’s approval before publication.

Availability: Available for guided evaluation. Some AI and advanced workspace screens use sample behaviour; external AI providers need configuration. A demo is not proof of production readiness. Check what works today.

Next step

Bring one real workflow question.

Evaluate one active project with your team. We can use that starting point to assess fit, show the relevant interface and make the remaining work clear.

Start with a question, not a commitment. Use sample or anonymised information.

Full functional walkthrough

Follow one conversation to an owned task.

Actual local interface from the product review build (labelled “Operating Flow”). 20-second screen recording, no narration. The Ask answer is a mock response. Local real API; review tenant is empty unless explicitly stated.

A silent recording of the local review build. It does not show the decision-review step: try that in the interactive journey on the home page, or ask for it in a guided evaluation.

Request a guided evaluation