Part 5: Who wins and who loses? How different types of business could be impacted by AI tools.

How Generative AI will impact product engineering teams — Part 5

This is the fifth part of a six part series investigating how generative AI productivity tools aimed at developers, like Github Copilot, ChatGPT and Amazon CodeWhisperer might impact the structure of entire product engineering teams.

In Part 4, we explored:

The opportunities Generative AI presents for organisations during an economic downturn and diminished venture capital funding.
Three scenarios for companies to take advantage of higher developer productivity with the introduction of Generative AI productivity tools: investing for growth, cost-cutting, and maintaining current budgets
Hypothetical consequences and challenges, like drastically altered team compositions and what these changes could mean for product managers and engineers.

Cui Bono — Who benefits?

There are always winners and losers when the business environment changes rapidly. So far, we’ve been discussing the benefits and the possible impact of Generative AI tools from the perspective of a hypothetical model, but we have glossed over a lot of unspoken caveats.

In reality, even if the benefits do play out as I’ve described in the past four articles, we’re unlikely to see the structural changes we’ve been discussing happen overnight.

I can attest to the fact that changing the type and number of roles in a business leads to many sleepless nights. Most organisations will simply choose not to address these changes directly because transformation is hard and emotional. The changes for most companies will take time to happen, slowly, organically and avoiding the need to make tough decisions. Sadly, that also means that many organisations won’t see the benefit that these changes could bring as quickly as they probably should.

Having considered how teams might benefit from these AI coding tools, I’d now like to turn my attention to how they may impact different types of organisation; newly minted startups with heady growth plans, smaller but stable companies with product and tech teams of less than a hundred people, larger organisations with tech teams of hundreds of engineers, and one group that I think will see particularly large impact: development outsourcing & offshoring firms.

Startups

Startups who are just receiving their first injection of angel or venture cash stand to benefit the most from a reduction of engineers required per product.

Just as with the arrival of Amazon AWS and the other hypercloud providers 10 years ago, the impact of AI coding assistants (and low code tools) will be a further democratisation (and demystification) of the ability to create web, mobile and internal back office applications. Put simply, it will be easier for less skilled teams to ship products to paying customers.

Even 15 years ago, the capital cost of buying servers and the expense of multi-year contracts for hosting services meant that the price of entry to launch a new product was prohibitively high for small businesses. With the growth of the cloud, entrepreneurs could scale on demand, with only a credit card (or free hosting credits) to their name. Today, the largest barrier to entry for founders with a product vision is the people cost of engineering, and these new tools stand to substantially reduce that cost.

Today in most VC backed startups, the technology is considered to be a large part of the unique value of the business. There is a trap that many founding teams fall into, that building as much of the tech stack as possible increases the value of the business. To make matters worse, this is sometimes exacerbated by some venture capital firms who press their young portfolio companies for more vertical integration and a de facto ‘build over buy’ mentality that prioritises intellectual property ownership. There is a cachet to be seen to have a company full of tech genius which can prove irresistible for both founders and investors.

However, for pragmatic tech co-founders, there is a big opportunity to reduce burn (the rate at which cash reserves are spent) and cut one of the highest expenses in the business — engineer salaries. According to this report by Carta, engineers typically make up 25% of the total headcount and 30% of total salary spend across their clients.

By choosing to focus heavily on optimising the use of both Generative AI Coding assistants like Copilot, but also choosing tools which minimise ‘clever’ engineering, like AWS Amplify Studio, Retool, Microsoft Power Apps and even integrations with point solutions, like Intercom’s Fin, founding teams can move at much greater speed than those that choose to prioritise the value of their intellectual property.

The generative AI coding tools we’ve been discussing in this series rapidly contract the gap between a product manager’s vision and the technical effort to implement that vision.

Smart startup founders in 2023 will start from a position of asking which tools they should be using to build their software, not whether they should be using generative AI tools. As these types of company benefit from having very little invested in existing teams, embedded technical biases and little legacy software to support, they will be able to take advantage of new skills and team structures from their inception.

For startups, generative AI tools will lower the cost and time to market for equivalent quality products, and also likely encourage them to have a far higher propensity to experiment, pivoting more quickly than more traditional peers.

Small businesses

For small but stable and established businesses that have two or three product engineering teams, there will be a good opportunity to benefit from more productive engineers. Companies that choose to augment their teams with generative AI tools should see higher quality code, more automation, shorter development cycles and more cost effective development.

With small and manageable development teams of 20–40 engineers, many of whom will have been with them for a number of years, these businesses are less likely to choose to reduce the number of staff. It will, however, be attractive to deliver more for the same cost.

Where cost reduction is necessary, the most obvious place for smaller companies to look will be to any outsourced development partners, especially those who are more traditional “bodyshopping” firms who are providing engineering, test or other technical skills. In these smaller businesses, if permanent engineers can be made more productive, the easier call will be to cut outsourced and more expensive day rates on contract roles than to cut internally.

(There will likely be a change in business mode for the outsourcers themselves, which we’ll come back to later)

For small businesses, there should be a small but positive economic impact as their smaller teams automate more and see higher quality outputs, with a relatively low risk of headcount reduction.

Larger companies

Larger businesses with tech and product teams of over a hundred people face a difficult choice in their long range planning — do they want to grow for the same cost, or maintain the same rate of product development for a substantially lower one?

Planning in these organisations is in years, not in quarters. Most will have long term relationships with a number of different outsourcing companies, as well as an internal mix of permanent and temporary technical staff. As we saw from our simple modelling before, the potential savings for companies of this size that choose to fully embrace generative AI tools are millions of £/$/€ each year.

However, although the financial benefits are significant, change in these companies is also hard. Established ways of thinking and of working, existing hierarchies, internal politics, the complexity of organisational change and the very significant cost of large scale redundancies will mean that planning is detailed and slow. While these companies consider how the changes may play out, it is likely that their teams will start to embrace tools themselves. As we’ve seen with other easy to use tech tools it will often be non-technical (but tech adjacent) teams like Marketing, that drive adoption. Inside product engineering teams themselves, there will be some drive to innovate and adopt tools, but understandably little effort to reduce the number of roles in the tech organisation.

For these large businesses, the stakes are high but change will be slow, and the benefits of productivity or cost savings will be slow to play out. However, for some enlightened and decisive organisations, the impact could be significant.

Outsourcing firms

Ah, I’ve been waiting for this one.

I’d like to say first, that there are genuinely great reasons to outsource, which I often break down as:

Engaging high value skills that wouldn’t be consistently used by the company
Adding flexibility to a stable base of permanent team members
Adding a capability that doesn’t differentiate your business

What’s not such a great use of outsourcing is to have large, multi-year contracts for relatively expensive resources that remain very consistent over time. Unfortunately, it’s still common, and a lot of outsourcing companies make a lot of money once they are part of the furniture of their clients.

Stability AI CEO Emad Mostaque recently stated in an interview that he believes that “outsourced coders up to level three programmers will be gone in the next year or two” and that “there will be no programmers in five years”. I think it’s fair to say that I’m less bullish that Mostaque, but still see enormous challenges for outsourcing companies in the near future.

For consulting firms that focus on delivering short term value, there is a clear and present opportunity for them to double down on the investigation of Generative AI tools, as they have the cash and incentive to resource their own teams to experiment with the tools and demonstrate best practice to their clients. There will be a temporarily rich seam, similar to the ‘Digital Transformation’ trend of 5–10 years ago, where these firms can offer consulting to help client companies adopt the new tools, offering training, frameworks and guardrails for their use. This will be especially comforting for more risk averse clients who will gain some peace of mind if they are receiving thoughtful counsel on Intellectual Property, security and data privacy issues that might otherwise be too significant for them to swallow.

As I highlighted before, any companies who believe that Generative AI tools should deliver cost savings will likely look at their outsourcing and offshoring costs before considering reducing the number of permanent employees.

Were I a CTO in a larger organisation confronted with the capability of tools like Copilot and CodeWhisperer, and learned that four out of five of my developers could be replaced with AI tooling, my initial reaction would be to start considering all of my outsourcing spend. In order to protect my high value internal engineers, it would be a relatively easy philosophical decision to slash the funding going to oursourcing firms.

It will be interesting to see this play out. Where outsourcing firms are providing highly specialised consulting services there will be a lower impact, but in the case that clients are outsourcing to take advantage of technical resources in lower cost jurisdictions, there will be a very real and growing impact.

Smaller boutique firms, who specialise in areas like mobile app development, web application development, QA or DevOps will be impacted differently than the giants like Accenture, Deloitte, TCS, Infosys and Wipro.

Those that should be the most worried by the advent of Generative AI tools are those suppliers that have ongoing bodyshopping contracts to provide fairly generic technical resources to clients, rather than those that partner to deliver specific outcomes or highly specialist skills.

In Part 6, we pass the baton to you with a request for you to experiment, and a reflection on the key things for early adopters to consider.

Read Part 6 here

Other articles in this series:

P.S. If you’re enjoying these articles on teams, check out my Teamcraft podcast, where my co-host Andrew Maclaren and I talk to guests about what makes teams work.