Note: This newsletter is a continuation of a prior one, “The Research Pulse,” in which I would surface new & useful working papers for a group of strategy & research professionals. I’ve decided to move it to Substack, and to address more timely and important topics.
Anyone trying to build a company today will tell you: to get anywhere meaningful, you need to be at the bleeding edge of available technologies in your most critical business functions. For our company, that’s places like - software development, sales & marketing, SEO, and content development.
Over time, I’ve come to recognize that each area (including the academic research market we sell into) follows a pattern in terms of adoption of AI and a host of other emerging technologies. I strongly believe that pattern is coming next for corporate research - namely: market research, customer research, and competitive intelligence.
Why elaborate on this? Simply because the potential is pretty astonishing if you can work through the initial up-front investments of time and get your teams going with the right mentality.
I mostly refer to it as the “hacking-tinkering paradigm” - working title. What it looks & feels like in terms of day-to-day work:
Testing out a huge number of products on a regular basis. And I mean, constantly. An unprecedented chunk of new tools are available B2C and with freemiums.
Getting precise about which software to use for which task. Model selection, search engine selection, which LLM for which type of work.
Building a tech stack of 5-7 products that talk to each other rather than relying on one tool for everything. Evaluating + tweaking all the time.
Copying and pasting between disconnected systems constantly. Or even taking the time to build webhooks + integrations - easier to do than ever.
Treating your workflow as a system you're always tinkering with. A lot more time spent towards the beginning of an initiative building out your processes.
A crucial theme in this evolution is an intense “verticalization” of AI tools. That is, a move from using the big LLMs like ChatGPT for one-off projects, to very specific applications - for instance, ones that have a set of “skills” relevant to your tasks, or ones that connect to and get updates from a particular data source of yours.
Yes, it’s a lot to stay on top of. And will involve sacrificing some immediate tasks. But staying on top of it is increasingly part of the job. It’s about building well-tailored workflows + tech stacks, and constantly looking for creative ways to improve them - which often requires not technological, but detailed functional knowledge.
Basic research itself, including information literacy, is unbelievably important in this new environment. You need to find new ways to, for instance, source competitive pricing data. Where to start? New tools? What should be the search? What sites to rely on? How can we demo and evaluate a handful of new tools quick & effectively?
There are three broad categories where most of these technological and organizational changes are happening:
Market research - recently, I ran market research reports in Perplexity to estimate TAMs, CAGRs, and key strategic findings mirroring two different reports that we had created in the past ourselves. The results were very, very good. Not perfect, but good. Hence, the most mechanical aspects of market research (e.g., consolidating research reports for a type of meta-analysis) are very likely to be more technology-enabled in the coming years. Coloop is focused specifically on this use case. On the organizational front, expect much more of this analysis to come out of shared services centers - indeed, we know from our discussions this shift is already underway.
Customer research - a lot of talk here about (1) improving the survey + interview process and (2) creating synthetic personas. The names I keep coming across are Listen Labs and, believe it or not, a tool just called Synthetic Users. The surveying tools appear to focus on developing huge networks of respondents; creating and deploying surveys to those respondents in highly customized fashion (e.g., with auto follow-up questions based on their responses); and synthesizing and reporting on results with AI. The synthetic personas are focused on using partial data on a customer segment to do things like simulate willingness-to-pay estimates. They are all the rage… but take it all with a grain of salt. A major Columbia study published this year found they barely outperform asking a raw LLM with zero customer data.
Competitive intelligence - the real innovation on this front comes down to a dirty word: scraping. I can recall working for several months on a project while at EY - specifically to prepare an Excel database with all the client engagements our competitors (e.g., IBM, Deloitte) had announced in the blockchain & cryptocurrencies arena. Doing so today would be (literally) free, 99% automated, recurring, and would probably take around 1 hour of set-up time. Currently, with our software business, we use tools like Apify to scrape data on our competitors’ social media posts, then feed that data into Claude via API to create reports. Looking at our subscription now: scraping for 1,000 results costs 40 cents.
Even if you have research or data challenges that can’t be placed into one of these buckets, or solved substantially with some magical new tool (indeed, many can’t), keep in mind one of the main underlying advancements of the big models like ChatGPT and Claude: personalized search, with a huge amount of context.
So even if you can’t use a fancy new startup tool to, say, track and analyze volume changes for very specific product categories, you certainly should be taking a good hard look at the databases and reports you use currently - because both finding and quickly evaluating new sources is as much a part of the story in this new paradigm as is landing on the perfect tool.
Tweaking systems > rifling through projects.
My hypothesis is that there’s a good opportunity to add value for research teams by simply staying on top of all of these changes, very intimately understanding the strengths and weaknesses of the tools available (with no affiliations - especially important given how much content in this space is sponsored / influencer-driven), and sharing insights regularly to the community.
I’ll continue my research, to monitor publicly available cases, and to speak with folks in industry about internal organizational changes they’re seeing.
I plan to open-source a lot of what I find. But also coming up with an offering that probably involves (1) a discovery process, (2) a very broad demo’ing & showcasing of available technologies, (3) sharing of detailed comparable cases and benchmarks, and (4) support with implementation, from hands-on training to helping secure internal buy-in.
Let me know any questions. Thanks all for reading.
-John
CEO of Sourced Economics + Moara.io; Adjunct at GW + Stevenson