The AI Revolution is Different in Speed and Nature to the Internet Revolution

Ray Poynter, 14 August 2025

---

As somebody who was a pioneer during the uptake of the Internet by the research ecosystem, I have been struck by how different I think the AI impact will be.

My feeling is that, compared with the Internet, the AI changes will be:

1. Faster
2. Driven by the big players
3. Take senior as well as junior jobs

The Internet Revolution
The Internet was launched in 1991, and market researchers started using it around 1994 and 1995. Online panels, such as Greenfield and Harris, emerged, offering a cheaper and faster method of reaching people. New agencies, like Pete Comley’s Virtual Surveys, were created to capitalise on this new medium. In the late 1990s, there was a deluge of Conference papers and other material. One example was the Chapter I wrote with Karlan Witt for the ESOMAR Handbook of Market and Opinion Research (1 Sep 1998) “Research on the Internet”, a guide to online surveys. This was despite the relatively small proportion of people who were using the Internet. In 1995, just 9% of the USA population was online. By 2000, the percentage of the USA online had reached 43% and the global penetration was about 6%.

One of the most noticeable features of the adoption of online surveys and online qual was that small agencies and innovative start-ups drove it. The larger, global agencies were much slower to adopt online.

By 2010, online research had become the global leader in terms of data collection modes, outpacing CATI and Face-to-Face in all the developed economies. 2010 was also the year that Wiley published my book “The Handbook of Online and Social Media Research”.

The shift to the Internet had several significant consequences for insights:

1. Research became much faster and cheaper, but it was not as good.
2. Lots of people lost their jobs (hundreds of thousands, possibly over a million), these were interviewers (F2F and CATI), questionnaire printers, data punchers etc. A lot of work has moved from high-cost countries to lower-cost ones (IMO, this is not a good or bad thing; it is simply a change).
3. It stopped the practice of approximating to random probability, high-quality samples, and moved us to online panels with panel members doing very large numbers of surveys.
4. It led to the growth of ResTech, which facilitated the development of in-house research. We now see, according to ESOMAR, about 50% of all research (by volume) being conducted internally by end-clients.

The AI Revolution
It is early days, but I think it is clear that if we take the launch to the public of ChatGPT in November 2022 as the start date, the growth of AI in research will be much faster than the growth in use of the Internet was.

This time, we are seeing the big companies, Ipsos, Kantar, Dynata, Toluna and the rest leading the charge to adopt AI. This has two implications for the speed of adoption: they have bigger budgets, and they are in contact with more end-clients.

The job losses this time are likely to be at all levels, from junior to senior. If I were being cynical, that is why market researchers seem more worried this time, they are in the firing line, not just the interviewers and clerical staff.

Similarities with the Internet Revolution

I do see a couple of key similarities with what we saw from the mid 1990s to 2010.

1. The driving forces will be speed and cost, and I would expect quality, on average, to fall.
2. I anticipate that more research will be conducted internally for end clients, with a further shift within these clients from insight teams to hands-on managers, designers, and C-suite executives.

Strategic Recommendations
Here are my key thoughts

1. Be the most AI-smart person in your team. If there is going to be a wrecking ball hitting your organisation, it is better to be driving it than to be on the receiving end.
2. Use AI to make you faster and ideally better. In the wider marketplace, faster and cheaper will be key. But within an organisation, being faster and better tends to lead to success.
3. At the moment, focus on using Agents; they are likely to be key in 2025 and probably 2026.
4. Look out for tips and tricks to be more effective. Generally, ignore people who say X can’t be done with AI; they may be correct (today), but it does not help you move forward. Look for advice on how to achieve things, how to minimise risks, and how to speed up your learning.
5. You will tend to help your organisation best by helping yourself by futureproofing your career. If you advance, your organisation has the chance to benefit too.