Just over a third (75.3%) of developers and tech leads believe that artificial intelligence (AI) has enabled them to make a quicker customer impact. Despite its potential power, though, there isn’t a blueprint for how businesses can successfully integrate this technology into their operations.
To help tech leaders navigate this potentially tricky evolution, OfferZen Engineering Manager Jason Tame chatted with Merelda Wu, Co-Founder and CEO of Melio AI, and Dries Cronje, Founder and CEO of Deep Learning Café, about how to strategise AI integration, build an AI-ready team and measure the success of your initiatives.
Disclaimer: AI has the power to reduce repetitive tasks that slow your team down, but it won’t be of much utility if you’re not in a position to use it effectively. Before kicking off your AI initiative, it’s important to determine whether your team has the capacity to take these projects on.
Align your AI ambitions with your business goals
Creating fit-for-purpose AI solutions starts with ensuring that your proposed solution will help you to achieve your overall business goals. According to Dries, there’s no need to complicate how you gauge the level of alignment.
“It’s not about looking for ways to implement AI in your business, but rather looking at what you need and seeing whether you can use AI to address that in your business in a way that aligns strategically with your business,” he said. And how might you do that? Dries added that you don’t have to reinvent the wheel to explore AI solutions:
“We still use the old-school SWOT analysis. We look at the strengths, weaknesses, opportunities and threats and where AI could fit into that paradigm to help you improve or overcome those things.”
Merelda agreed on this point, noting that tech leaders must identify their top business problems, ideate potential solutions and create a realistic roadmap for how the AI implementation could play out.
“You have to identify your business goals and bring in the technology lens to figure out how long it’s going to take, how much it’s going to cost and whether you’re ready to jump in. Otherwise, a lot of the time, these initiatives end up just fizzling out and it leaves your software engineers in a sad state,” she said.
Assess your AI readiness
The first step in driving a successful AI implementation is ensuring that your organisation and team are actually ready for it. According to Merelda, there are three areas you need to assess to establish whether AI could work for your operation.
1. Business readiness
The first is business readiness. You can ask yourself a few questions here:
- Can you link your AI initiatives to your business goals?
- Do you know what your baseline is?
- Do you have time and money to support your AI use cases?.
2. Data readiness
The second thing you want to assess is data readiness. Here you can ask yourself:
- Can you find the answer that you’re looking for with existing data that you already have?
- Do you know if the AI will perform better or faster than a human at doing this?
3. Integrated readiness
Lastly, you can look at technology or integration readiness. Here’s the main question Merelda suggests you ask yourself:
- Do you already have a mature software development team using cloud technologies?
If you have a mature software development team, outsource the AI build to a trusted vendor and bring it in-house once the MVP is complete.
If you don’t, look for managed service providers or AI-as-a-Service to operationalise it for you.
And of course, if you have an AI team, let them guide you on buying versus building.
Weigh up the benefits and costs
Another important factor to consider, especially for smaller businesses with tighter budgets, is how quickly your team can implement AI in your operations.
“For small to medium-sized companies, the speed of implementation is probably the biggest factor for AI success. If you can’t roll it out fast enough, then you’ll incur so much technical debt that it might not make sense,” noted Merelda.
Here, you’ll want to consider what tools or technologies can drive a faster speed of implementation or potentially consider cloud-based solutions.
“A great tip for really small companies is to find other small companies with similar problems that you can pitch together with to implement AI in your businesses. Then you can all use the solution and you can all contribute towards making it better instead of trying to do it yourself,” added Dries.
Build a team of AI enthusiasts
Your team will play a key part in ensuring that your AI implementation is successful. Whether you’re building one from scratch or shifting existing roles to focus on this tech, you need to be sure that your developers have the capacity to do this work.
"It's challenging to ask a busy team to take on learning new technologies outside of their usual tasks," said Merelda
"The key is to integrate AI experimentation into their existing projects, allowing them to explore new tools while contributing to their current work. It's important to create a culture where they can experiment, enjoy the process, and feel safe failing – because each failure will contribute to building your company's collective intelligence, strengthening your ability to tackle future AI projects."
Naturally, tech leaders play an important role here too. They need to educate everyone within the business about the possibilities of AI, get developers excited about the transition to AI and cultivate the culture of problem-solving within the company. Dries suggested two practical ways to get everyone in the company involved:
- Create a channel for AI: it’s a place where everyone can share information, articles and updates about what they’ve been doing with AI.
- Create opportunities where you can build together as a team: whether it’s once a month or maybe every couple of weeks – dedicate an afternoon or a day to a hackathon or do a one-day sprint where we can build something together as a team.
Measuring the success of AI initiatives
There isn’t much point in implementing AI in your operations if it doesn’t bring a benefit for your business. To determine whether it’s doing what you envisioned, you need to understand where you’re starting from and set key performance indicators (KPIs).
“You need to figure out what the technical and business success criteria are. Technical success could be the accuracy, latency or throughput of the models. Business success criteria are a little bit more vague, but it might be something like a conversion rate or the number of hours spent to solve a problem,” Merelda explained.
“Your baseline is very important. To measure your return on investment you need to know how you’re currently performing and you need to continuously measure the same metrics once you’re in production.”
Dries noted that identifying and tracking your success across these metrics is also important because of the multiplier effect this can have.
“With a bit of success, you breed more success and you create excitement around your project. This makes it easy, especially in larger organisations, for the project owner to go back to the business and communicate the success and get investment for future projects,” he said.
The bottom line
There are plenty of opportunities to integrate AI into your business operations, like improving your interview process, evaluating developer skills or increasing productivity.
However, you’ll need to strategically assess your business needs and readiness, engage your team and set clear benchmarks for success if you want to ensure that your AI initiatives are not only feasible, but also impactful.