Many companies are adopting artificial intelligence (AI) to drive strategic decisions or introduce new business models. According to a global McKinsey study, 56 percent of all respondents report AI adoption in at least one function in 2021, up from 50 percent in the previous year.

Whether your company is just getting started in its AI journey or you are leveraging it across multiple business functions, there are numerous competitive advantages to gain from improved employee experience, deeper customer insight, and enhanced business functions.

Human-first AI

Many companies starting out in AI make the mistake of focusing too early on which AI technologies to use or which data to collect. In reality, your people are the key to a successful AI program. The people in your business must understand the KPIs that are important and how to collect the right kind of data. Using this, data scientists can make use of AI tools to provide business insights and predictions.

One of AI’s biggest challenges today is the potential for systems to introduce bias because the teams that built them were not sufficiently diverse. AI systems rely on data and inputs and diverse AI teams are necessary to avoid bias.

[ Check out our primer on 10 key artificial intelligence terms for IT and business leaders: Cheat sheet: AI glossary. ]

Create a well-rounded team that includes members with diverse backgrounds and leverage data from various groups and sources to ensure your models perform well externally. Ethics and monitoring must remain top of mind as this technology continues to evolve.

The three pillars of AI

• The data: Data is a valuable commodity and an important part of any AI system, so ensure you have lots to work with, emphasizing quality rather than quantity.
• The objective: For a successful AI system, ask yourself what AI can achieve and the business objective.
• The model: This is essentially programming. The model takes historical data and historical outcomes to predict the target based on real-time data.

Each pillar plays an important role. You need data to train a model, but you also need a model to observe data – it’s a classic chicken-and-egg problem. When introducing AI, start small, implement systems and tools, and staff the right people to unlock this potential.

Your first hire should be a data analyst and a data engineer to ensure that data keeps flowing. Leverage turnkey analytics and business intelligence tools to easily uncover correlations and help your data speak.

Once you have these three things set in motion and the staff to support it, you can create a business that is built on AI and scale from there.

AI is an experimental science​

With any experiment, you often don’t know what you don’t know. In the case of AI, intuition is useful, but it can be dangerous if you rely too heavily on it. Keep an open mind and constantly test, measure, and keep records of experiments.​​

Have an ongoing conversation internally about ethics, biases, and discrimination. On our team, we created an AI ethics community to ensure we do not introduce unexpected biases and that AI is used for good. Allow your employees to drive this and collaborate. People are passionate about ethical AI and want peace of mind that technology is used for the right reasons.

Additionally, ensure your technology is compliant. Work closely with local regulatory bodies to understand what you can and should do and consider personal data (GDPR, consent management) and accountability.

[ Related read: Building a learning culture with AI ]

Aligning AI with your business​

There is no one-size-fits-all solution for AI; each use case is unique. To stay competitive, first assess your business: What areas are most important, what can be streamlined, and how can you best use your employees’ time? This information will help determine where AI fits in your workflow.

It can be hard to meet a target goal that achieves your ultimate business objective. Instead, start small and follow a crawl-walk-run approach by working on smaller, more achievable targets.

Early in the Criteo journey, for example, our AI focused on predicting whether an end user would click on an advertisement – but for retailers, the end objective is sales. While a user clicking on an advertisement was not a direct predictor of sales, we could use it to predict accurately and it led to the true target.

Continue to refine these proxies and implement more accurate models as your AI evolves. As we became more proficient in predicting, we moved to more complex business targets, such as sales, revenue, or even margin.

AI should enable better experiences for your employees, leadership, clients, and partners. If you’re just starting to leverage AI, consider what areas will have the most impact and where you have the most data, and then scale from there.

Test, analyze, and test some more

If you have intuition, test it: Build a solid infrastructure that empowers your team to try new ideas, test quickly, and roll back easily if needed. Consistently monitor your AI and ensure you have a backup solution in place, especially if you are in a life-impacting space.

AI is a fast-paced, dynamic field, and organizations that stand still will fall behind. Invest in your program and keep a pulse on new enhancements and innovations. Several years ago, Criteo invested in its AI Lab. Some people questioned the decision, but today it powers our products through published research, product innovations, and new technologies. This research helps blacklist inappropriate products from advertising banners and enables advertisers to more accurately predict if users will click on an ad.

If your organization is early in the AI journey, start small but start early to stay competitive. Embrace collaboration and participate in open source by leveraging resources and contributing. Most importantly, remember to focus on your people – your employees, clients, partners, and customers. Take a human-first approach to determine how AI can be a force for good and solve business challenges.

[ Want best practices for AI workloads? Get the eBook: Top considerations for building a production-ready AI/ML environment. ]