Carrol Chang says Andela has specialized in software engineering for more than a decade.
“software engineering is the profession and the job that Andela has specialized in for our entire decade plus history.”
Person
Carol Chang
1 episodes · 21 quoted claims
Carrol Chang describes agentic coding as engineers giving directions, guardrails and context to AI coding agents that execute the work, rather than writing every character of code themselves.
“rather than handwriting code, writing every character, what they are doing now is they are giving directions and guidance and guardrails and some context to AI, and then coding agents are executing the work.”
Carrol Chang says AI can do software engineering work at a level once expected of entry-level workers, though not at a senior or perfect level.
“it's capable of doing the job at a breathtaking level of competency that you would have typically thought of an entry level worker needing to learn”
Carrol Chang defines hard skills as technical skills such as programming languages like Python and Java, and frameworks like .NET.
“There are hard skills, and these are the things that might sound like Python or Java or net”
Carrol Chang calls problem solving, strategic thinking and judgment 'squishy skills' and says they are now at a tremendous premium.
“I've heard them called squishy skills before. These are things like problem solving, strategic thinking, judgment, and that bucket of skills is now at a tremendous premium.”
Carrol Chang says AI is making knowledge of hard technical skills less important.
“what AI is doing is it is making people's knowledge of all of the hard skills less important.”
Andela's hiring process gives candidates a prompt and data and a week to prepare a final presentation, which is treated as a live conversation.
“We give them a prompt, we give them some data, and then we ask them to take a week and to come back and deliver a presentation.”
Carrol Chang says recent candidate presentations of five or ten pages have been hard to read because so much was obviously AI-written.
“those five or ten pages are incredibly hard to read because so much of it is so obviously written by AI”
Andela invested heavily in technical assessment over several years because it runs a large global network of engineers.
“we invested a lot over the last several years into assessment because we run this very large global talent network of engineers.”
Carrol Chang says candidates in unguided coding tests will likely paste the task into AI tools such as Claude Code, Codex or GitHub Copilot and return the output.
“people are gonna just go and put the prompt into Claude Code or Codex or GitHub Copilot and then take the outputs and and give them back to you”
Andela built an assessment system that assumes and encourages candidates to use AI.
“So we actually built a system in which we assume and encourage that people are gonna use AI, because that's actually how the job is changing anyway.”
Carrol Chang says Andela can detect answers that merely reproduce a coding agent's output and gives them no credit.
“we can spot that, and they'll effectively be scored a zero. They'll get no credit for that answer.”
Carrol Chang says that once someone has strong judgment and problem solving, learning a missing technical skill is usually the easier part.
“usually then learning that hard skill is not the hard part.”
Carrol Chang says Andela was founded as a learning-focused company.
“Andela is a company that was founded as a learning focused company.”
Andela follows the 70-20-10 learning model, under which about 70% of learning comes from doing the work.
“We believe in the seventy, twenty, ten learning model, which says that 70%, meaning the vast majority of one's learning comes from actually doing the work.”
Carrol Chang says that when AI tools are rolled out, at least about 5% of employees, from any department, become enthusiastic early adopters.
“once you unleash AI tools at a company, you will always have 5% or more of these folks who are enthusiastic early adopters.”
Carrol Chang warns that using individual AI-skills assessments to monitor workers loses their trust and can make an AI rollout backfire.
“If you're assessing everyone individually in the workforce just because you're trying to do a monitoring exercise, right? That's not gonna win the trust of the workforce.”
Carrol Chang predicts that in five years successful companies will have many AI workers at the bottom of the organization handling rote, repeatable work.
“It's gonna look like an org that has a lot of AI workers at the bottom, and those AI workers are gonna be in charge of a lot of rote, tedious, repeatable and automatable work”
Carrol Chang relays that an Andela customer had her best quarterly business review because AI meant engineering was no longer the bottleneck.
“I had my best QBR ever, or I had my best QBR of my last year just recently. Because now engineering is no longer the bottleneck, and AI is what's enabling this.”
Carrol Chang says AI is starting to write product requirements documents (PRDs) and engineering requirements documents (ERDs), which could move the bottleneck beyond product.
“write a bunch of PRDs and ERDs, the product requirements documents and the engineering requirement documents.”
Carrol Chang predicts that within five years teams will be more global and distributed because AI widens access to knowledge work.
“teams are gonna look a lot more global and a lot more distributed than they are today”