The Cost of Making Something Easier

When somebody says a tool saves time, I want to know whose time it saves.

That question is the beginning of an estimate.

Imagine a business using software to draft replies to customer questions. The first reply becomes quicker to produce. Someone still has to check what it promises, decide whether it fits the situation and deal with the customer when the reply is wrong. (This is a hypothetical example; whether the arrangement helps depends on how those pieces actually work.)

The saving might be large. It might also be a transfer: one person spends less time drafting while another spends more time detecting small, persuasive mistakes. We should be able to describe both possibilities without turning the conversation into a referendum on civilization.

Look outside the first task

A workflow has more edges than the screen in the demonstration.

There is the person whose work is being changed. The colleague who will check it. The customer who receives it. The person responsible when nobody can explain why the system did what it did. There may also be people whose work helped produce the tool but who never appear in the purchase conversation.

I would put those people into the discussion early, as a practical question rather than a list of stakeholders in a proposal: what changes for each of them?

Does the tool remove tedious work, or remove the part of the work through which someone learns to recognize mistakes? Can an employee challenge the output? Can a customer reach a person? Is the organization keeping enough knowledge to operate when the service is unavailable or the commercial arrangement changes?

The answers do not have to point in one direction. A useful tool can create a new responsibility. A decision can be worth making while still costing something you would rather have kept.

A preference is information

A business owner may tell us that they do not want generative AI in their software. I treat that as part of discovering what they need.

We can talk about the reason. Privacy might matter. So might the effects on employees, creative work, reliability or dependence on a provider. The person may simply want a system they can understand and operate without that component. We can design around the requirement.

It helps to make the requirement precise. Software with no generative AI in its operation is different from software built without AI assistance at any point. The second is a restriction on our development process as well as on the finished system, so it changes the estimate. Neither should be hidden inside a vague assurance that everything is "AI-free".

Clarity gives both parties something they can actually agree to.

Put the trade-off in the proposal

For a particular project, I would rather see a small, specific account of the choices than a broad statement of ethical ambition.

What information enters the system? Where does it go? Who may inspect it? What is proposed, and what is performed automatically? Who checks the result, and do they have the time and authority to do so? What happens when the tool is declined, disconnected or wrong?

Ownership and licensing questions need attention too. They should be raised as actual questions for the project and the relevant specialists, and a sales sentence shouldn't settle them. Resource use belongs in the discussion where it can be meaningfully assessed; a vague green claim tells the buyer very little.

No single number makes every consequence comparable. We can still make the consequences more visible and decide which ones the people involved are willing to accept.

Keep the ability to choose

My preference is for systems that leave people with more practical options. That can mean a carefully bounded use of AI. It can mean ordinary software. It can mean deciding that the current process should be changed before any new tool is introduced.

We are willing to investigate all three, and to explain why one seems a better fit for the particular situation.

Making something easier can be a wonderful thing. It is worth checking that the difficulty has gone away, and hasn't simply moved to someone who was not in the meeting.