SMRIT LABZ  —  Rooted in Wisdom, Driven by Innovation.

Approach

People stay in charge: how we think about automation

At SMRIT LABZ, intelligent systems do the repetitive work and people make the decisions that carry consequences. The principle, and how it shapes what we build.

SMRIT LABZ · · 4 minute read

A business owner pausing to read a draft on a tablet before approving it

Intelligent systems can now draft a letter, summarise a report, write a social media post or answer a routine question in seconds. That is genuinely useful. It also raises a question every organisation has to answer: which decisions should a system make on its own, and which should always come back to a person?

At SMRIT LABZ our answer is a simple principle: systems do the work; people make the decisions that carry consequences. This note explains what that means and how it shapes what we build.

Why the question matters

When a system acts on its own in an organisation’s name, three things are at stake.

  • Accuracy. Automated systems can be confidently wrong. They can state a fact that is out of date, misread a request, or fill a gap with something plausible but untrue.
  • Accountability. When something goes wrong, someone has to answer for it: to a customer, a regulator, or the public. “The system did it” is not an answer anyone accepts.
  • Trust. People trust an organisation because of how it behaves. A single careless message, sent automatically, can undo years of careful work.

None of these are reasons to avoid automation. They are reasons to be deliberate about where a person stays in the loop.

Where we draw the line

We sort the work into three kinds.

Work the system can do on its own. Repetitive, low-risk tasks where an error is easy to see and easy to fix: preparing a draft, sorting incoming messages, checking that information matches a source, reminding someone of a deadline, collecting figures into a report.

Work the system prepares and a person decides. Anything that goes out in the organisation’s name, commits money, or affects a customer: a public post, a reply to a complaint, an advertising plan, a change to records. The system does the preparation, often most of the effort, and a person reads it and decides.

Work that stays with people. Judgement that depends on context the system does not have, or on values: how to respond to a sensitive situation, whether a claim is fair, what the organisation stands for.

The boundary between the first two is a choice the organisation should make, not the system. In one of our products, for example, a business can choose to let a few simple replies go out on their own, such as a thank-you to a positive comment, and can switch that off at any time. Posts, pictures and advertising plans always wait for a person.

What it looks like in practice

The principle turns into concrete design decisions:

  • Nothing goes out without an approval. Drafts wait in a clear queue. Approving is a deliberate action, not a default.
  • A record of who decided. Every approval records who made it and when, so an organisation can always answer “who approved this?”
  • Easy to say no, or to ask for a change. A person can send a draft back with a sentence of instruction, and the system redoes it.
  • Warnings, not silent blocks. Where we can detect a likely problem, such as wording that breaks an industry’s advertising rules, the system says so before approval, in plain words, and the person decides with that information in front of them.
  • Facts from the record. Systems should write from information the organisation has confirmed, such as its address, hours and services, and not invent details to fill a gap.
  • Approval that fits the day. Decisions have to be easy to make where people actually are, including from a phone, or the approval step becomes the bottleneck and people start working around it.

The trade-off, honestly

Keeping a person in the loop is slower than letting a system act alone. A post waits for an owner to look at it; a reply waits for someone to read it. We think that trade is worth making for work that carries consequences, and we try to reduce its cost by making the decision itself quick: the draft is ready, the context is beside it, and approving takes a tap.

It also means the responsibility stays where it belongs. The organisation decides what it says and does. The system makes it faster to do well.

Where this applies

This principle runs through everything we build, across the industries we work in, from healthcare and education to infrastructure and enterprise. It is most visible in our first product, Baila Lento, a digital marketing platform in which every post, picture and advertising plan waits for the business’s approval before it goes out.

If you are thinking about where automation fits in your own organisation and where it should not, we are happy to talk it through. Write to us at connect@smritlabz.com.

More insights