Artificial intelligence is accelerating an enormous number of processes inside companies.
And I'm not only talking about administrative processes.
Today we can use tools like ChatGPT, Claude, or Gemini to analyze information, draft documents, summarize meetings, review emails, generate reports, and help out in practically every area of the business.
But there's something I personally still prefer not to fully delegate to AI:
The commitments we make with our clients.
And there's a very simple reason.
When a company signs a document, a quote, a commercial proposal, or a project scope, it isn't simply approving a piece of text.
It's taking on commitments.
And those commitments can have economic consequences.
The Document That Can Change a Project's Profitability
In our company, when we present a commercial proposal for a project, we usually have a document we call the project scope.
In that document we detail what we're going to do.
For example:
- Requirements gathering.
- Process analysis.
- Process design.
- Process optimization.
- Feature development.
- Configuration.
- Integrations.
- Testing.
- Management activities.
- Deliverables.
- Client and provider responsibilities.
All of that represents commitments.
That's why, even though we use AI for many other activities, we review that document manually.
It may even take me longer.
But I'd rather invest that time before signing.
Because if I'm not aware of what I'm committing to, I might discover the problem months later.
And by then it may be too late.
The Problem I'm Seeing in Some Companies
One thing that has caught my attention in recent years is how some executives are using AI to review documentation.
It has happened to me with clients.
I send a proposal or project scope document and expect the person responsible to review it.
But instead of reading it carefully, marking observations, and discussing what isn't clear, they do something different:
They attach the PDF to ChatGPT, Claude, Gemini, or whatever AI tool they use and ask:
"What do you think of this document?"
Or:
"Does it meet these criteria?"
Or even:
"Is it fulfilling the project requirements?"
And sometimes the process ends there.
The person receives the AI's answer and considers the document reviewed.
I've even received client responses that have practically the same structure as the AI-generated answer.
And that's where the real problem appears.
The person thinks they reviewed the document, when in reality they delegated the review.
The Problem Shows Up Months Later
Let's say we're executing a project.
We reach a certain stage and the client tells us:
"But that feature was supposed to be included."
So we review the scope.
And we find that it wasn't.
We show them the commercial proposal.
We show them the document they approved.
We show them the signed version.
And sure enough, that feature or activity wasn't included.
The problem is that the executive probably thought it was.
Why?
Because they didn't do a detailed review of the document.
AI had reviewed it for them.
And here I want to make an important clarification.
It doesn't necessarily mean the AI gave an "incorrect" answer.
It may even have done an excellent summary.
The problem is different:
The final decision was made by a person who didn't necessarily read and understand what they were approving.
AI Doesn't Know Your Business Commitment
This point is fundamental.
An AI can analyze a document and tell you:
"The document appears to meet these requirements."
But that doesn't necessarily mean it can take responsibility for the decision you're making.
AI doesn't know all the commercial consequences of your approval.
It doesn't know how much it will cost you to execute an additional activity.
It doesn't know what resources you'll have to hire.
It doesn't know whether a certain phrase could be interpreted differently during project execution.
It doesn't know what relationship you have with that client.
And above all, it isn't the one who will have to execute the commitment.
You are.
What About the Context Window?
There's also a technical point to understand here.
AI models work with a context window, which determines the amount of information the model can consider within an interaction.
Current context windows can be huge and have improved a lot.
That's why it would be wrong to simply say "AI can't read long documents."
It can.
The problem is more subtle.
When a person feeds large amounts of information—a contract, several appendices, emails, specifications, conversations, and other documents—they shouldn't assume that an AI-generated answer automatically equals a thorough contractual or business review.
Depending on the tool and the workflow used, there can be interpretation issues, contradictory information, details that don't get enough attention, or conclusions that seem reasonable but need to be verified.
And there's an even more important problem:
AI can help you find something, but that doesn't mean you should leave it the responsibility of deciding for you.
AI Adoption Is Already Massive
This isn't a hypothetical problem.
AI is already part of the daily work of millions of professionals.
According to the Stack Overflow Developer Survey 2025, 84% of respondents use or plan to use AI tools in their development process. Additionally, 51% of professional developers reported using them daily.
But there's an important difference between using AI as an assistant and using it as a substitute for professional judgment.
The same study found that 46% of developers distrust the accuracy of AI tools' answers, compared to 33% who trust them.
That's a significant signal.
Even people who work directly with technology recognize that AI answers need validation.
And this shouldn't surprise us.
AI is a tool.
It isn't the project's decision-maker.
The Mistake Isn't Using AI
Let me be very clear about this.
I'm not saying you shouldn't use ChatGPT, Claude, Gemini, or other tools to review documents.
In fact, they can be extremely useful.
You can ask them to:
- Summarize a contract.
- Identify important dates.
- Find contradictions.
- Compare two versions.
- Flag topics that should be reviewed.
- Extract obligations.
- Organize requirements.
- Search for specific terms.
- Generate a list of questions.
- Compare a proposal against a set of criteria.
All of that can save you a lot of time.
The problem appears when we move from:
"Help me review this document."
to:
"Review this document for me and tell me if I can approve it."
Those are completely different things.
A Good Rule for Executives
I have a very simple recommendation for business owners and executives.
When someone on your team presents a proposal, a quote, a contract, a project scope, or any document that involves an important decision, don't settle for an email reply alone.
Ask them to explain verbally:
What do you understand we're approving?
Then ask:
What are our commitments?
What's out of scope?
How much could this cost us to execute?
What risks did you identify?
Which part of the document raises doubts for you?
The conversation can quickly reveal whether the person actually reviewed the document or simply asked an AI to do it.
And this is especially important when we're talking about technology projects.
Because many times the document that looks like "a few pages" can contain commitments that later represent weeks or months of work.
AI as an Assistant, Not the Decision-Maker
I believe this will be one of the great challenges for companies in the coming years.
It won't just be about learning to use AI.
It will be about learning when to use it and when not to delegate a decision to it.
AI can read.
It can summarize.
It can compare.
It can analyze.
It can find patterns.
It can help us enormously.
But when we're talking about commercial commitments, economic decisions, contracts, project scopes, or responsibilities toward a client, we still need something no tool can fully replace:
human judgment and responsibility.
So my recommendation is simple:
Use AI to work faster.
Use it to find problems.
Use it as a second pair of eyes.
But when you're about to take on a commitment with a client, read the document yourself too.
Because if you sign the document, the responsibility won't be ChatGPT's.
It will be yours.
AI can help you read the document, but it shouldn't be the one deciding what you're committing to.
If you want to review the architecture, scope, or security of a software project before taking on a commitment, talk to a specialist.
Sources:
