Pakistan AI

Stop counting AI users. Start measuring AI literacy

Every survey counts users. Almost nobody measures whether people can tell when the machine is wrong. That gap decides whether AI makes Pakistan's firms better or merely louder.

MEA Tech Watch graphic: opinion by Ali Tahir on measuring AI literacy among marketers in Pakistan
Graphic: MEA Tech Watch
Advertisement

Top Banner

For the past year I have been studying AI literacy among marketers in Pakistan‘s IT sector as part of my doctoral research. The results are not published yet, so you will not find them here. What I can share is what a year of reading, fieldwork and arguing with my own assumptions has done to the question I started with.

I started out wanting to know how much marketers use AI. I now think that is the least useful question anyone can ask. Every survey, every board deck and every national policy counts users. Almost nobody measures literacy, and literacy is the thing that decides whether AI produces new work or just faster versions of the old work.

Use is the wrong number

Look at the numbers the industry quotes. Microsoft’s 2026 Work Trend Index surveyed 20,000 full-time employees who use AI at work across ten countries. Two thirds reported higher efficiency and 58% said AI let them produce work they could not have produced a year earlier. In the same study, 26% said their leadership was clear and consistent about AI, and 13% said they were rewarded for reinventing their work with it.

Advertisement

In-Article

Read those together and the picture is of a workforce that has adopted a tool faster than anyone has taught it what the tool is for. Use went up. Understanding was never measured. “Did you open the tool this week” is a question about habit. It tells you nothing about whether the person can spot the paragraph the tool got wrong.

In marketing that gap is expensive. Ours is the department where fluent output goes straight to a customer. A confident, well-formatted, wrong claim about a product does more damage from a marketer than from almost anyone else in the building, and generative tools are very good at confident, well-formatted and wrong.

What literacy actually means

The definition most researchers work from comes from Duri Long and Brian Magerko at Georgia Tech, who in 2020 described AI literacy as a set of competencies that let people critically evaluate AI, collaborate with it and use it as a tool at work. The word that matters in that sentence is evaluate. Literacy is not knowing which button to press. It is knowing when the output is not to be trusted, which tool fits which job, and where the ethical and privacy lines sit.

That is a different skill from prompting, and it is the one I see missing from most training I come across in Pakistan and the Gulf. Courses teach people to get more out of the machine. Very few teach people to doubt it. In my own work on go-to-market teams across several markets, the people who got real value from these tools were not the heaviest users. They were the ones who treated every output as a first draft from a bright, overconfident intern.

Why Pakistan’s IT sector is the place to watch

Pakistan’s IT and telecom exports reached a record $4.6bn in the year to June 2026, up about 21%, according to State Bank data. The marketers inside those firms are selling to buyers in the United States, the United Kingdom and the Gulf, often against competitors with ten times the budget. AI is the great equaliser in that fight, and also the fastest way to embarrass yourself in front of a procurement committee in Texas.

The government’s National AI Policy, approved in July 2025, sets a target of one million trained AI professionals by 2030, with 3,000 scholarships a year. I welcome the ambition. My worry is the definition. If “AI professional” means an engineer who builds models, the policy will miss the far larger group of people who will decide whether AI makes Pakistani firms better or merely louder: the marketers, salespeople and account managers who put its output in front of paying customers every day. A million people trained in tools and not in judgment is a million people who can be wrong at scale.

Advertisement

In Article – 2

Even the regulators do not measure it

Europe is the one place that has written AI literacy into law. Article 4 of the EU AI Act has required employers to take measures to ensure “a sufficient level of AI literacy” among their staff since 2 February 2025, with enforcement from 2 August 2026. Then comes the quiet part: the European Commission’s own guidance says there is no obligation to measure employees’ AI literacy levels. Firms are asked to provide training and keep a record of it. Nobody has to check whether it worked.

I understand why. Measuring a competence is harder than counting a course. But it leaves the law exactly where most companies already are: a line in the training budget and no idea what it bought. For Pakistani IT firms with European clients it is also a coming question. Sooner or later a client will ask how you ensure the literacy of the people touching their data and their brand, and “we gave everyone a ChatGPT licence” will not be an answer.

The case against me

The strongest objection is that literacy is a researcher’s word for a problem that solves itself. The tools change every month, the argument goes, so teach people to prompt, let them learn by doing, and the judgment will follow. I used to half believe this. I no longer do, for one reason: tool knowledge is perishable and judgment is not. A prompt trick I learned last year is already useless. The habit of asking “how would I know if this is wrong” will outlast every model on the market. If you only have budget for one, buy the one that does not expire.

What I would do

To the chief executive of an IT exporter: before you buy another licence, measure. There are short, validated instruments in the published literature that take ten minutes to run. Find out whether your marketing team can tell an AI feature from an ordinary one, choose between tools, and judge an output. Then spend on the gaps, not on the seats.

To marketing leaders: make evaluation a job task, not a value. One rule I would put in every team tomorrow is that no AI-assisted claim about a product, a price or a competitor goes to a customer without a named second reader. It costs minutes and it builds the exact muscle literacy describes.

To the Ministry of IT and the Pakistan Software Export Board: when you count towards the million, count literacy in non-technical roles, and publish how you measure it. A target nobody can audit is a slogan.

To the universities: put AI literacy in the marketing curriculum as a competence to be assessed, not a module to be attended. The students graduating in 2027 will be the first generation for whom this is simply part of the job.

When my research is published I will write about what it found. Until then I would rather leave you with the question than a premature answer, because the question is the useful part. Do you know, today, who on your team can tell when the machine is wrong?

Topics
Advertisement

End Article

Ali Tahir Contributor Ali Tahir contributes analysis on telecom, payments and the business of technology. He has spent more than 17 years in B2B marketing and communications across fintech, software, logistics and ICT, in Pakistan, the Gulf, the UK and the US. He is also a doctoral researcher studying how B2B companies adopt AI and marketing technology. More from Ali Tahir

Keep reading