MUMBAI, India, October 13, 2026 : Is AI taking entry-level jobs in financial markets? Respondents to a QuantInsti survey are divided. Asked whether the entry-level role is disappearing or changing shape, respondents answered:
● 25.8% said it will have fewer seats
● 30.3% said it will change shape, with roughly the same number of seats
● 32.6% said it is too early to say
● 11.2% said it will grow
The survey was run during and after QuantInsti's 16th Anniversary Annual Algo Trading Conference, held on September 24, 2026, which drew participants from more than 110 countries. It is voluntary and self-selected, so it reflects the views of the 185 people who chose to respond, not all conference participants. The question did not ask respondents to attribute any change to AI specifically. What follows is how respondents answered the questions people in trading communities keep asking.
Are entry-level roles disappearing?
Views differ by experience. Share of respondents in each group who expect fewer entry-level seats:
● Under 1 year in markets: 24.4%
● 1 to 3 years: 39.5% (the highest of any group)
● 3 to 7 years: 17.8%
● 7 to 15 years: 15.8% (a smaller group)
● 15 or more years: 25.0%
Mridul Arora, Partner at Elevation Capital, speaking on the conference's careers panel, said the worst-case view is too strong, though he did not dismiss the change.
“Has AI closed the door for entry-level people? I think that's too dramatic. I feel like some of the old doors are closed.”
Mridul Arora
He also noted that firms are building differently: “People are building for leverage and not for headcounts.”
What skills do respondents expect to matter?
Each respondent chose the single skill they expect to matter most for getting hired over the next two years:
● 49.2%: getting real work out of AI tools
● 14.9%: statistics and validation methods
● 14.9%: market and product knowledge
● 13.3%: communication with stakeholders
● 7.7%: writing and reading code
Few respondents chose coding as the single most important skill, which does not mean they see it as unimportant. The survey was not designed to measure that. About 78% of respondents are self-directed traders or students rather than hiring managers, so the results reflect community expectations, not employer data.
Ajay Sharma, President of Edelweiss Public Alternatives, drew a line between screen-based tasks and those that depend on judgment.
“Anything that happens on screen, looking at charts, placing a trade, a computer and AI will do better than me. But anything in the physical world, which is soft skills, you cannot replace with AI.”
Ajay Sharma
Do you need to code to trade systematically?
Most respondents do not yet. Asked how much of their trading runs through code:
● 35.2% said none of it, everything is discretionary
● 24.7% are not currently trading
● 23.6% use code for research and place orders manually
● 11.0% run automated execution under supervision
● 5.5% are fully automated
In total, 40.1% use code in some way (the last three groups combined).
Coding and AI use go together. Share of respondents who have not used AI in the past three months:
Respondents who already run code in their trading: 9.7%
Respondents who do not run code: 39.6% (a group that includes people not currently trading)
This shows an association and does not prove that coding leads to AI use.
Tony Lee, CFA, Chief Brokerage Officer at Alpaca, said the engineers his firm hires still need to code.
“They know how to code, but more importantly, they know how to prompt the agents to code. Their ability to communicate with the agents is the most important thing we're looking for nowadays.”
Tony Lee, CFA
Can AI build a trading strategy?
72.1% of respondents have used AI for some trading task in the past three months. Each respondent named their main use:
● 27.9%: strategy ideation or research
● 27.3%: writing or debugging code
● 9.8%: data pipeline work
● 7.1%: reading research faster
● 27.9%: have not used AI yet
Among those who have used AI, about nine in ten named something it produced that they would actually use, most often help with code or strategy ideas. 22.2% of respondents worry that everyone will end up with the same AI-generated ideas.
At the conference's opening workshop, Stefan Jansen, founder of Applied AI and author of Machine Learning for Trading, showed how coding agents built and tested a trading strategy in minutes, then walked attendees through how to check the result.
Why do good-looking backtests fail live?
A backtest tests a strategy on past data. Respondents' biggest concern about AI in trading was a strategy that looks strong there and then fails in live use:
● 38.9%: results that look good in a backtest and fail live
● 22.2%: everyone ending up with the same generated ideas
● 16.7%: their existing skills counting for less
● 16.1%: not being able to explain what the model did
● 6.1%: regulatory and compliance uncertainty
Among self-directed traders, 47.0% named the backtest problem as their top concern.
Jansen showed one way the check can go wrong. A “t-statistic” is a number researchers use to judge whether a strategy's success is a repeatable pattern or luck. In a recorded walkthrough presented during the workshop, a standard calculation for the AI-built strategy gave a strong-looking 6.1. After correcting for the fact that consecutive predictions overlap, it fell to 2.0, which is far less convincing.
“The naive t-statistic is 6.1. Corrected for the fact that consecutive labels overlap, it is 2.0. One of those numbers gets funded. The other one is true.”
Stefan Jansen, in the walkthrough
Where should someone start?
Sameer Kumar, Chief Technology Officer and Partner at iRage, said domain knowledge comes first.
“You still need to talk the same language to the AI. You need to have the jargon... Get the understanding, get the basics right, and then you should delve in.”
Sameer Kumar
Arora's advice for people retraining was practical.
“Build with AI. Let that project you build be your CV.”
Mridul Arora
What the survey shows for India
Just over half of respondents were based in India, and the comparisons below are against smaller groups elsewhere. India compared with the Americas and Europe/UK:
● Have not used AI in their trading workflow: 35.7% in India, against 24.2% in the Americas and 21.4% in Europe/UK
● Named writing or debugging code as their main use of AI: 16.3% in India, against 45.5% and 39.3%
● Named backtest failure as their top concern: 27.4% in India, against 57.6% and 39.3%
● Expect fewer entry-level seats: 21.1% in India, against 42.4% and 17.9%
The takeaway
Respondents' answers do not settle whether AI is taking entry-level jobs. A third say it is too early to tell, and the rest are split. What they describe is AI being added to a set of skills, including statistics, domain knowledge and communication, that they still expect to matter. The full survey report has the complete results by role, experience and region.
About QuantInsti
QuantInsti Quantitative Learning Private Limited
QuantInsti is a pioneer in algorithmic and quantitative trading education, having trained finance and trading professionals across 150+ countries. As the creator of EPAT® (Executive Programme in Algorithmic Trading), one of the world's most recognised programmes in the field, QuantInsti combines deep domain expertise with cutting-edge curriculum design. QuantInsti recently launched programmes focused on agentic AI applications in trading, keeping its learners at the forefront of the industry's most transformative developments. Its team of industry practitioners, researchers, and pedagogy experts is committed to delivering measurable learning outcomes that translate directly into professional impact.
For more information, visit: https://www.quantinsti.com/
