Consumers today are more AI-savvy than you think. Customers use their best interaction experience in one domain as a baseline expectation in others. This means that, when it comes to AI, every single business is in competition with the global giants, including Amazon and Netflix. AI is no longer a nice-to-have feature; it’s a must-have — and poor AI has become a real threat to businesses.

When the algorithm fails to live up to people’s expectations of the user experience and doesn’t deliver the service its users want, the people using the system end up feeling annoyed, frustrated, and tired. New research my team was involved in has identified this phenomenon, and it’s hurting businesses everywhere.

Consumers’ patience is running out. We are at a tipping point where companies in many industries need to step up and build better AI to survive. Indeed, AI is no longer just about the technology; it’s about how customers experience your brand.

In my 15 years working as a consultant for digital businesses, I’ve worked to resolve many of the pain points that come with corporations implementing AI. Here are four things I’ve learned about realigning any organization to build better AI.

1. Focus where it counts

One of our clients — a Nordic data-intensive public authority serving millions of people — had been successful in deploying various forms of AI-powered chatbots in customer interactions across their operations. While their customers appreciated these chatbots, the chatbots soon turned out to be a bit of a distraction for the business. The most significant opportunity lay instead in applying AI to the company’s own back-office operations, the service that ultimately produced the real customer value.

This called for a different, more complex type of cooperation within the organization. The project required new cross-functional teams, and as the firm was not quite mature enough to support these efforts, AI implementation needed to be prioritized as a top management issue, coordinated as a company-wide, top-down effort.

From a business strategy perspective, AI brings the most impact when it is used to create or optimize distinctive capabilities (sources of competitive advantage) instead of table-stakes (non-differentiating must-haves). Introducing AI into the core processes of your business is hard but it may also be critical if you want to change your competitive game.

AI development also requires constant iteration and it is best done in cross-functional teams. In other words, AI is a focused, company-wide effort. The idea of cross-functional teams for AI may seem simple and compatible with existing organizational structures, but in practice, that is seldom the case. Many attempts fail. AI must therefore be involved in business strategy work from the get-go: It should be recognized as a potential source of threats and opportunities in the business environment and acknowledged as a force that can influence the entire future trajectory of your organization.

2. Think long-term

A large European grocery chain we work with first started applying AI in marketing automation, a non-differentiating must-have. Soon, however, they realized the serious business possibilities with AI and began to view artificial intelligence as an indispensable in-house capability. Rather than attempt to calculate ROI for a specific business case, they saw hiring the right talent and building other key enablers for AI as an investment that would pay off over time through its impact on the core business. This took patience and foresight. The firm’s AI capabilities have since taken years to build, but it now has sophisticated AI handling assortment management, one of the distinctive capabilities and sources of competitiveness and profitability in the high-volume grocery business. The investment also better positioned the company for the future: Its AI capabilities have proven a key asset in the battle for overall market share brought about by the strong growth in grocery ecommerce.

Regardless of what you build or buy, or of the talent composition of your teams, it’s important to secure control of critical resources and ensure you build AI capabilities in-house over time. That means it’s crucial to reframe AI as an investment not as a cost. Organizational capabilities in AI may take years to build, but the business benefits are likely to become substantial over time. Once in place, AI capabilities can show a very high business yield — and can do so quickly. Even large companies can become agile to the point where new AI applications can be created in as little as one week.

3. Loosen the reins

A global industrial company we work with has been quite successful in building AI applications on a local business line or product level using cross-functional teams. The problem, however, is that these AI capabilities are now scattered across the group in pockets of excellence. This works well on a local level, but it also means a lot of potential and efficiency are lost due to the lack of learning and capability development across the units. To then scale the business impact of AI while retaining the strong business anchoring calls for building out a centralized mechanism for developing general AI capabilities. The aim is not to centralize control of application development but to best support such development in an overarching, coordinated way.

Purposeful AI development requires both direction and degrees of freedom. Cross-functional teams for AI must stay focused and dedicated but must also have enough autonomy to carry out the exploration and development needed to build AI. There has to be a continuous awareness of what the desired business outcomes are but also an openness to explore the possibilities and detours that inevitably come up in all AI exploration and development.

Organizing around AI is about balancing managed expertise and local application. The hub and spokes approach is one way to enable cross-functional teams at scale while still retaining control of the general capability. The optimal setup is a balance between a centralized and a distributed approach. If your AI capabilities are too centralized, they end up detached from the business; if they’re too distributed, they fail to create impact. You should organize AI development with a hub-and-spokes model, according to the overall AI maturity of your organization, and make sure you balance business goals with your teams’ freedom to experiment.

4. Show, don’t tell

A large luxury fashion retail group we worked with wanted to bring in advanced analytics across all its core business processes, including marketing, purchasing, merchandising, and pricing. The initiative faced fierce resistance in some functions that operated on human experience and intuition. We built ground-level AI solutions that helped doubters at their own trade and visibly improved their performance. One such solution was a machine learning–based segmentation tool that revealed a novel, clear distinction between brand loyalists and the customers who expected to be served with variety. Our solution also demonstrated to the stakeholders the benefits of working with such segmentation. This then significantly helped build understanding and buy-in for the overall agenda.

Building with AI is to a large extent about creating the right organizational mindset. AI solutions aim to augment or replace human cognitive tasks and decision-making, and such intrusions on human intuition typically face varying degrees and forms of resistance in an organization. One way to overcome this is to let your people see the realized benefits of AI for themselves, in their own work. AI can streamline processes, free up time and space, and help make the work more strategic and interesting. When people are involved in the process and are gradually shown these realized benefits of AI, they tend to become invested in making sure that AI flourishes in the organization. From there on, the rest then tends to fall into place.

Olof Hoverfält is a leading Strategy & Business Design expert at the technology consultancy Reaktor.