EXCLUSIVE: “Buy, Build, or Boost: Choosing the Right Path Amidst an Accelerating Agentic Future” – Matthew Barnard, BBD and Theodora Lau , Unconventional Ventures in ‘The Fintech Magazine’

AI is fundamentally reshaping the buy vs build decision in financial services. With genAI tools and open source models flooding the market, we caught up with BBD Software’s Matthew Barnard and finfluencer and author Theodora Lau to discuss governance and differentiation in an agentic world

To buy, or not to buy? That is the question facing the world’s most ambitious enterprises as they consider how best to implement the innovations that will enhance their business model.

With artificial intelligence (AI) rewiring software development, should organisations be building their own systems or buying in existing solutions?

There’s an expansive spectrum of options now available as the technology matures. If a company chooses a reputable off-the-shelf offer, they benefit from a proven tool they can deploy at speed and at scale. If they build a bespoke solution – provided they have the right talent – then they might produce a fully customised, controlled, and future-proofed answer to their needs. But, increasingly, enterprises are adopting a hybrid approach, working with specialist providers to develop flexible, scalable, and expert-driven solutions.

One such is BBD Software (BBD), which partners with clients from multiple sectors, including financial services and insurance, to help them design, build, and deploy AI-led systems, guiding them from experimental proofs of concept to launching solutions in production environments. With 40 years of experience in software engineering and digital transformation, BBD provides nearshore and offshore development teams that help enterprises navigate the AI arena.

“What we’re seeing with our clients is that everybody’s experimenting, everybody’s learning, and everybody’s growing,” says Matthew Barnard, Executive Director at BBD. “The difficulty is that what you learned two weeks ago has already been replaced, and will be replaced again in two weeks’ time. But you can’t be a bystander on this journey; you need to get involved with it.”

There is no right or wrong answer: the decision to buy or build depends on context and the regulatory environment in which a business operates. For many, a successful future will depend on finding the right balance.

“There is a lot we can do with modern AI tools,” says Theodora Lau of Unconventional Ventures, a thought leader in this space and the author of Banking On (Artificial) Intelligence: Navigating The Realities Of AI In Financial Services. “We can write code, we can prototype much faster and easier. But, at the end of the day, when you’re shipping something out, there’s complexity behind it. It’s that we still need to think about.”

Counting the comparative costs

AI is starting to have a profound impact on how financial services are delivered to consumers, and the technology has undoubtedly helped lower barriers to entry when it comes to building software solutions. Indeed, it’s often no longer a case of creating from scratch; it’s a question of customising one of the plethora of open-source models in the market.

Vibe coding – where instead of writing code line-by-line, developers (novices and the more experienced alike) use natural language to prompt AI assistants like ChatGPT to generate, debug, and refine applications – has become a buzzword in 2026. It enables rapid prototyping and iteration, helping companies test ideas cheaply and quickly without accruing high initial development costs.

It’s changing the dynamic. What was once perhaps considered too expensive or time-consuming to build is now seen as more achievable. Projects previously earmarked for software-as-a-service (SaaS) providers are being pulled back in-house and reassessed as realistic builds. Many banks and financial institutions still rely on the legacy programming COBOL (Common Business Oriented Language), first designed back in 1959, specifically for business data processing. Barnard explains how vibe coding is making COBOL upgrades and modernisation much more accessible for in-house teams.

“Those projects would have typically taken years to write new code to replace them,” he explains. “And while they still take a reasonable amount of time, they’ve become much more viable. Some of the estimates we’re putting in to clients are half the price of what they would have been three years ago, so business cases are certainly shifting at the enterprise level.”

Lau believes this shift will necessitate changes in how providers think about costing their services. “There will need to be more flexibility when it comes to pricing,” she says.

“The old way won’t work well anymore; there will need to be a rethink focussed on outcomes.”

Barnard agrees about the need for flexibility, recognising that buying solutions comes with compromises. Organisations often only use a fraction of the functionality while paying for the full product.

“I think there’s a place for both buying and building, and always will be,” he adds. “People often buy in SaaS software for the 40 or 50 per cent of the functionality they want but they’re paying for everything.

“Now the business case for building is becoming more cost-effective, providers need to find new ways to enable people to pay an appropriate fraction of the licence fee to get just what they need.”

The true value of a consultative partner

Custom builds may provide the ability to tailor solutions precisely to specific business needs, but they also introduce challenges encompassing the complexity of bringing software into production, long-term maintenance, and governance.

While cost and convenience are important, speed alone isn’t a silver bullet. Enterprise environments – particularly in the multi-layered financial services arena – require much more than rapid development and deployment. This is where the distinction between building and buying becomes more nuanced.

The more robust off-the-shelf solutions offer maturity, embedded expertise, and proven functionality, representing years of intellectual property (IP) and refinement that help enterprises overcome challenges.

“For small, simple applications, vibe coding and that sort of thing is fine,” says Barnard. “But financial markets are deep. There’s a lot of knowledge embedded within SaaS companies; a lot of IP behind them, and complexity that’s built into that software.

“It’s not the speed of coding that makes a difference – it’s that IP and knowledge that gets it working well, covering elements like accuracy and security. That can’t just be pulled out of thin air. So there’s a level of maturity that’s needed to allow enterprises to start taking advantage of the AI models out there.”

Accuracy is a point that Lau also picks up on. In financial services, systems must be reliable and precise under pressure.

“You’re dealing with people’s money, and you’re dealing with regulators,” she says. “You can’t be 90 per cent right, you have to be 100 per cent. Also, bad actors will always go where the money goes. If you just vibe code and ship everything to production without understanding how everything ties together, particularly from a cybersecurity perspective, then you’re going to get into trouble.”

Driving differentiation

The conversation also underlines an important division in terms of where differentiation occurs, and how customisation can help drive loyalty, whether implemented in-house or in tandem with a partner. Barnard and Lau both believe that core systems, especially in heavily regulated areas like payments and accounting, tend to remain standardised, with differentiation increasingly seen in the customer-facing front end. AI is helping to reimagine interface design and how services are delivered, allowing enterprises to personalise and adapt experiences more quickly.

“Enterprises should think about where they see the technology as an advantage,” says Barnard. “They’re looking to do something unique that their competitors haven’t done, and that’s more often in the front-end interaction with customers rather than in the back-end accounting systems, which are based on international standards.”

“A good place to start is the cultural perspective,” adds Lau. “Financial institutions operating across different regions can trial how to draw on different languages and linguistic features to create an experience that feels personable to those they’re serving.

“Call centres and bots are a good area to explore. I came across one example where AI is used to detect where the caller originates from and changes the accent of the agent to foster familiarisation. That’s a little bit of a slippery slope, but it’s interesting what’s out there.”

Such innovation starts from automating simple tasks that aren’t business critical, and iterating from there so that the organisation can build knowledge and momentum.

“If you look at people who are successfully putting things together quickly, they get the first 90 per cent done, and then the last 10 per cent takes the bulk of the time because the devil is in the details,” says Barnard. “How does it run in production? How does it run at scale? And how are you able to support and maintain it? Those parts are key in finance.”

In Banking On (Artificial) Intelligence: Navigating The Realities Of AI In Financial Services, Lau argues that while AI is becoming essential for innovation and efficiency in customer service, its adoption must always prioritise responsible, human-centric design.

She explains how AI can be harnessed to improve lending processes through iteration.

“Start small,” she says. “Get all the documents you need together so the AI can do all the back-end tasks that previously required stacks of paperwork.

“Make sure you have a human in the loop, checking the outputs. Make sure you can trace it back to why a decision was made.”

Opening up to an agentic future

Looking ahead, one of the most significant trends to consider, whether you’re building or buying, is the acceleration towards agent-based systems and more structured enterprise AI adoption. According to Gartner, by 2028, AI agents will be responsible for delivering 50 per cent of enterprise architecture outcomes, improving the quality and consistency of strategic architectural decision-making. This shift – which Gartner refers to as ‘augmented enterprise architecture’ – will move human enterprise architects away from manual execution and toward strategic advisory and critical AI oversight.

Cutting through the jargon, what this means in practice is that rather than relying on ad hoc vibe coding, organisations will need to formalise how AI is integrated into their development processes, focussing on accuracy, governance, and repeatability. This is prompting greater regulatory scrutiny, with ongoing debate around risk, control, and transparency.

Firms need to be aware of different approaches in different jurisdictions. The European Union’s AI Act, for example, supervises the use of AI in financial services by classifying systems like credit scoring and risk assessment as ‘high-risk’, imposing strict obligations that include mandatory risk management, human oversight, and post-market monitoring. Compliance is required of both providers and users.

The US, meanwhile, lacks a single, comprehensive AI law, relying instead on sector-specific regulators, like the Securities and Exchange Commission (SEC), using existing consumer protection laws. And the UK is currently leaning into a context-based approach rather than strict upfront restrictions, focussing on empowering existing regulators to apply these principles.

“I wouldn’t be surprised to see more demands from regulators,” says Lau. “They’ll want to know what companies are doing, and they’ll want to see audit trails. They’ll expect firms to be able to explain outcomes and how they’re mitigating any risks that are introduced.

“As we move towards building agentic commerce, there are things in the back end that we need to start thinking carefully about. Do we know the agent that’s initiating the purchase? Is it a valid agent? What kind of scope and abilities is the agent allowed to adhere to? And how do you validate it?”

With so much noise and hype surrounding AI, Barnard concludes that it’s difficult to know exactly what the future holds.

“It’s hard to look 12 or 18 months ahead because things are changing every month,” he says. “I think it’s best to take it quarter by quarter, and you need to be on that journey of actually learning and growing.

“At BBD, we’re investing our learning with our enterprise clients, exploring the possibilities of agentic AI. We’re creating agents with very specific skills and context, and testing how they can replicate the work being done by a human – whether a developer, a business analyst, or a project manager.

“And then we’re orchestrating those agents to do various tasks, and doing so carefully because, at the enterprise level, you can’t be nearly right – you have to be very right.”

This article was published in The Fintech Magazine Issue #38, Page 8-10

The post EXCLUSIVE: “Buy, Build, or Boost: Choosing the Right Path Amidst an Accelerating Agentic Future” – Matthew Barnard, BBD and Theodora Lau , Unconventional Ventures in ‘The Fintech Magazine’ appeared first on FF News | Fintech Finance.