Have you come across one of those adverts, whether posters or short clips, where so much information is compressed into such a small space that you do not know where to look?
They feature tiny fonts, several colours, emojis, generic illustrations, sometimes strange-looking people with six fingers, and enough information to fill a report squeezed onto one page.
Most of us immediately think: “Oh, this came from an AI prompt.”
Sometimes, you are left puzzled when you realise the advert comes from a real, respected institution: a national company, government agency, non-governmental organisation, university or even a multinational corporation.
The troubling part is not that AI was used. AI is here, and it can be incredibly useful. What worries me is when you can see that nobody went beyond the first prompt; nobody checked the information, corrected the illustration, reconsidered the layout or simply made the final product their own.
In other words, someone delegated not only the tedious work but also the judgement. And that, I think, is where our problem begins.
The digital age, now accelerated by generative AI, has changed how we produce and consume information. In much of the digital environment, dissemination is driven by “engagement,” so we increasingly communicate through short social-media posts, one-minute videos, catchy phrases, emojis and quick graphics. There is nothing inherently wrong with that; sometimes brevity is exactly what communication requires.
However, there is real danger in assuming that all knowledge must fit that format. Some things require explanation. Some things require nuance, and quantitative information definitely does.
More than numbers
If I tell you an interest rate is 10 per cent, the number itself tells you very little. Why 10 per cent and not 12 per cent? What assumptions produced it?
What happens if it increases to 15 per cent? Which group benefits, which group carries the risk, and is the result still sustainable under a different economic scenario?
READ MORE: Chatting With GPT: A Crocheter’s Observations
Those questions cannot always be reduced to one beautiful dashboard. Yet I increasingly encounter this in boardrooms.
For starters, any quant knows that you can make numbers tell the story you want without falsifying the underlying data. The choice of baseline, scale, aggregation or visualisation can change the story a reader sees, even when the data remain exactly the same.
Quantitative presentations have morphed into an endless stream of graphs, figures, charts and summary statistics, sometimes so compressed that they begin to feel like an alien language. We present numbers beautifully without necessarily communicating what those numbers mean.
In academia and higher education, the problem may be even worse. Sometimes academic experts, who are supposed to help solve society’s problems, stand before non-technical audiences and speak almost entirely in technical jargon.
Years ago, during my schooling days, my father told me something that stayed with me: an expert is not necessarily the person who can make a subject sound complicated. An expert understands the subject well enough to explain a complicated phenomenon in a way another person can understand.
Structure matters
That distinction matters.
A policymaker should not need a PhD in mathematics to understand a model’s implications. A board member should not need an actuarial qualification to understand why a stress test matters. Technical sophistication and understandable communication are not opposites.
As a quant, however, I must admit something upfront: first impressions matter, perhaps more than we academics like to admit.
READ MORE: No Policy, No Shield: Why Tanzania Is Still Unprepared for the AI Era
A report may contain brilliant research, but if the reader opens it and immediately meets inconsistent headings, long, boring paragraphs, badly positioned equations, figures floating far from their explanations, tables squeezed beyond the margins, and references formatted in five different ways, the presentation begins competing with the knowledge.
Writing is especially unforgiving because the reader cannot immediately stop you and ask, “What exactly do you mean here?”
In academia, reports, theses and dissertations are among the main ways we communicate research findings to fellow experts, industry and the wider public. Their structure matters, and the same applies to presentations.
The industry–academia relationship works properly only when knowledge can move in both directions: industry identifies challenges, and academia researches possible solutions; academia develops methods, and industry tests, adapts and applies them. Innovation happens somewhere in that exchange.
The LaTeX test
However, the exchange collapses if one side cannot understand the other. This is where formatting becomes more than decoration.
Formatting is the architecture of a document: typography, spacing, margins, numbering, citations, headings, tables, figures, equations and the general organisation of information. Good formatting should almost disappear; it should quietly guide the reader through the writer’s thinking.
For years, I corrected student reports. The technical work could be good, yet I repeatedly found myself correcting the same things: equation alignment, table placement, inconsistent numbering, citations, captions, spacing and document structure.
Eventually, I stopped. Instead, I required the reports in LaTeX. The students were not impressed. Their first reaction was predictable: why create extra work when the output was still just a report?
READ MORE: A Genuine Lever or More of the Same? Tanzania’s AI Education Guidelines at a Crossroads
From their point of view, they already knew word-processing software. Now I was asking them to learn commands before they could even write an equation. It looked like unnecessary labour imposed by a lecturer who happened to like mathematical typesetting.
So I made it compulsory. More importantly, I required them to submit the source files alongside the final PDF. There was no escaping it.
I did not, however, ask them to start from a completely empty screen. I provided templates based on the standard LaTeX book, report and article document classes for written reports, as well as Beamer for presentations.
Their job was to learn the commands and then discover the mathematical, graphical and actuarial packages that suited the work they wanted to produce. That was when things became interesting.
At first, they learnt only enough to survive the assignment. Then competition took over; granted, my students are quite competitive.
Once one group discovered how to produce a particularly elegant table, another wanted a better one. Someone discovered a package for a particular type of graph; someone else found a more interesting way to present an actuarial formula.
Colours appeared. Timelines improved. Tables became clearer, graphs became more intentional, and illustrations began to appear where they actually helped explain the mathematics.
Then sophistication followed.
One group discovered how to integrate work done in R into reproducible reports using R Markdown and LaTeX. Another figured out how to move work developed in Python through Jupyter notebooks into similar LaTeX-and-PDF workflows.
Essentially, they had started searching online for packages, tools and solutions that I had never taught them. That was my first pleasant surprise. The second was much more important: the writing itself became sharper.
The reports became more personal. You could read different groups and recognise different styles of thinking; even when several students worked within the same thematic area, the reports no longer looked like slightly altered versions of the same document.
One group might approach the problem historically. Another might begin with the data, build the discussion around a practical application or use a sequence of graphs to develop its argument gradually.
The subject was quantitative, but the students had started telling stories. That really changed how enjoyable the reports were to read.
A well-constructed 70-page quantitative report stopped feeling like 70 pages of equations, tables and model outputs. With my mathematical mind, some became almost as enjoyable to read as a novel because I could follow the writer’s train of thought: where the problem started, how the evidence developed, the analysis and calculations that followed, where uncertainty appeared and how the conclusion was reached.
READ MORE: Tanzania Court Adopts Artificial Intelligence (AI) in its Processes
Then came AI. Naturally, students discovered that generative-AI tools could help them write LaTeX commands, troubleshoot errors and locate packages. I had no problem with that.
In fact, this was one of the outcomes I liked most. Even students who used AI shortcuts increasingly used it in the way I think such tools are most useful: to handle repetitive or mechanical tasks and verify the result.
They could ask an AI assistant why a table refused to compile, how to resize a figure, which package supported a particular actuarial symbol or how to automate a repetitive command. However, they still had to decide what the report was saying, determine the order of the argument, choose which graph mattered, which result deserved emphasis, and how the pieces fit together.
AI became the assistant, not the author. This is the real lesson I took from the whole experiment.
Not an advertisement
I must put a disclaimer here: this is not an advertisement for LaTeX. Word-processing applications can produce excellent, professional documents when used properly, and LaTeX can equally produce a terrible document in the hands of someone with nothing meaningful to say.
A beautifully typeset bad argument is still a bad argument. What changed was not simply the software.
Students became more deliberate about constructing a document. Because they had to think about structure, commands, mathematical notation, tables, figures and presentation, they also became more conscious of how they communicated their ideas.
The process encouraged a certain discipline, and once that discipline became familiar, creativity followed. This matters far beyond quantitative disciplines.
Tanzania and Africa more broadly have substantial expertise in formal and informal sectors. We have researchers, engineers, doctors, economists, actuaries, statisticians, entrepreneurs, artisans, farmers and specialists with great practical knowledge.
Yet much of that knowledge remains within specialised groups. Any knowledge that cannot travel has limited impact; if experts cannot explain what they know outside their own circles, how will society apply it?
The challenge of the AI age, then, is not merely learning how to use increasingly powerful tools. It is learning how to use them without surrendering the skills that make knowledge human: judgement, curiosity, creativity, interpretation and the ability to tell another person, clearly, why it matters.
Therefore, the question is not whether we should use AI, LaTeX, Word, PowerPoint or whatever tool arrives next. We should. The more important question is whether the tool remains our assistant or quietly becomes our author.
We now live in an age when producing information has never been easier. Perhaps the skill we must protect most carefully is the ability to make that information mean something not only to us but also to the society around us.
Agnella Nemuo (PhD) is a quant and academic specialising in financial mathematics and actuarial science. She’s available at agnella.nemuo@gmail.com. The opinions expressed here are the writer’s own and do not necessarily reflect those of The Chanzo. If you are interested in publishing in this space, please contact our editors at editor@thechanzo.com.
