In just a few years, artificial intelligence (AI) has gone from an emerging technology to an essential one. AI companies raised a record $33 billion in equity funding in 2020, according to CB Insights data, and businesses that identify as AI extend beyond the tech sector. Many companies involved in more than 870 AI-related mergers, acquisitions and public offerings between 2016 and 2020 were in industries such as transportation, aerospace, healthcare, biotech, electronics and advertising, a PwC analysis of CB Insights data found.
In addition, more companies aspire to be “AI first,” upgrading their data infrastructure and adopting new formats for sourcing and storing data. They’re increasingly relying on AI models for everything from monitoring performance to developing ethical guidance that complies with new regulations.
But this increased interest raises new questions:
Before considering an AI deal, acquirers and other investors need to be clear on the rationale and have a rigorous process for evaluating the transaction. Anything less could limit the odds of turning AI’s potential into reality.
According to PwC’s 2021 AI predictions survey of US business and technology executives, most of today’s biggest tech companies are accelerating their embrace of AI. A quarter of the companies surveyed reported widespread adoption, compared with 18% a year earlier.
Many believe AI will improve productivity and enable them to operate more efficiently. In addition, they see the technology as improving the customer experience, enhancing cost savings and improving internal decision-making. AI systems capabilities continue to grow, broadening their appeal. For example, AI can now generate text, audio and images that most humans can’t distinguish from original material, and this will have a variety of applications in business messaging, customer service and training.
Whether companies intend to acquire talent, data or analytics, AI investments tend to fall into one of the following five categories:
The buyer and target join forces to conduct long-term research into new AI applications aimed to accelerate growth of the overall business. Such deals include Google-DeepMind, Intel-Moovit, Microsoft-Lobe, Microsoft-Softomotive, Oracle-DataFox, Workday-Stories.bi and Amazon-Zoox.
Buyer acquires virtual reality or other platforms to integrate and enhance existing AI capabilities, as demonstrated in transactions such as Facebook-Oculus, Snap-AI Factory, ServiceNow-Element AI, Apple-FlyBy Media and Amazon-Body Labs.
Buyer aims to enhance its product offering with specific technology that could be a market mover for them. Examples include NVIDIA-Mellanox, ServiceNow-Loom Systems, Cisco-ThousandEyes, Palo Alto Networks-Aporeto, Juniper Networks-Mist, Avalara-Indix and Amazon-Kiva.
Buyer acquires target for its AI talent or intellectual property developed through AI, as evident in such deals as Apple-Init.ai and ServiceNow-Element AI. Industry continues to appeal to graduating PhDs in AI, attracting 65% in 2019 compared with 44% in 2010. Companies are scrambling to find skilled talent needed to meet production goals.
Buyer acquires target for its analytics capabilities, such as in Microsoft- ADRM Software, Apple-Turi, Apple-Laserlike, Google-API.ai, Google-Terraform Labs, SAP-Qualtrics, Salesforce-Datorama, Facebook-GrocStyle Salesforce-Tableau and Accenture-Byte.
Before jumping into a deal, acquirers need to evaluate these transactions through a different lens, or else their investments won’t generate the value they expect. Deals involving AI aren’t always what they appear to be. According to a 2019 MMC Ventures study of more than 2,000 European AI startups, 40% were found to have no AI at all.
Because the nuances of AI can make each transaction unique, dealmakers should carefully evaluate the technology and the target before proceeding. Moreover, integration plans should consider responsible AI principles and practices to ensure a deal’s value continues long after it closes.
Buying AI capabilities rather than building them in-house has advantages, but only if companies can navigate the challenges. For example, companies that acquire AI technologies can deploy them in the market faster, but they may have less flexibility in designing a system tailored for their products or services. Acquiring AI also can help companies develop technical expertise quickly and avoid depending on third-party vendors for critical tools.
However, these benefits should be carefully weighed against the cost of the deal and potential pitfalls, such as culture gaps between acquirer and target, a lack of customization of the existing technology and the risk that the technology doesn’t perform as promised.
Understanding the capabilities and ensuring the proper evaluation methods are in place is particularly critical for non-tech companies aiming to enhance operational efficiency. How, for example, does a food service company determine whether a specific technology can improve order fulfillment?
Dealmakers need to think strategically and ask: Why is an AI company worth buying? What business problem will the technology solve?
Depending on a company’s business strategy, the reasons for acquiring AI capabilities commonly include:
For a deal to succeed, companies need a rigorous method for evaluating the transaction. The process should include input from financial experts, product managers, data scientists and technologists who can assess whether the target’s AI capabilities will be able to perform as promised. These diverse perspectives can evaluate deals from different angles and address questions such as:
AI doesn’t work without the right data, but targets often overstate the accessibility and uniqueness of these assets. Product managers, data scientists and technologists can contribute to commercial diligence efforts by assessing how the data compares with that of competitors.
This level of collaboration can pay off in big ways. For example, one tech company that acquired an AI-based cybersecurity company found that the target’s technology wasn’t as differentiated as it would have been a few years earlier. The buyer uncovered the shortcoming through an integrated diligence process, with the product technology and commercial diligence teams working together to interview customers and evaluate the commercial and competitive aspects of the target.
Companies should consider responsible AI in determining if a transaction will reach its optimal value. This starts with an integration plan that accounts for the technology’s evolving sophistication. A plan should not only outline when and how the assets and operations of an acquirer and target would combine, but also cover how the new company will respond to regulations and social standards likely to shape the future of AI.
The rise of AI raises inherent issues about trust and accountability to customers, governments and other stakeholders. For example, what happens if your AI makes decisions that are systematically unfair to certain groups? Numerous studies have found that AI can contain racial, gender and disability biases. The technology may be called artificial intelligence, but it still depends on people to write the algorithms, choose the data the algorithms use and apply the results.
Companies acquiring AI capabilities need an integrated governance system to mitigate these risks and identify potential biases before they can contaminate results. Here are some principles and practices of responsible AI that companies should consider:
Dealmakers should coordinate with the acquirer’s and target’s compliance departments and other stakeholders to address existing and developing regulations. Take, for instance, privacy laws. As an acquirer and target combine key data that powers the AI, they must assess whether the systems comply with Europe’s General Data Protection Regulation (GDPR), the California Consumer Privacy Act (CCPA) and other relevant privacy laws. If not, the acquirer, target or both should consider anonymizing data or taking other steps to mitigate noncompliance issues. In addition to existing legislation, acquirers should also examine the Institute of Electrical and Electronics Engineers’ (IEEE) and European Commission’s respective guidance on AI ethics.
AI systems need to be secure and safe. The value of a deal involving AI often hinges on the notion that the data being acquired has not been tampered with and is protected. As a result, dealmakers should evaluate how vulnerable a target’s AI systems are to cyber attacks and whether proper controls are in place.This starts with involving chief information security officers in the deal early on, so security experts can help a deal team understand potential synergies with the target environment. Risks differ depending on the industry. Companies should evaluate these risks as well as the potential financial, operational and reputational impact.
At some point, a company may need to explain why a particular AI model reached a particular decision. Dealmakers should rely on the target’s management, data scientists and others with knowledge of the models to address this question and tailor the explanation to different stakeholders, including regulators, business sponsors and consumers. Doing so inspires confidence among employees and consumers in the AI technology acquired and can further protect the acquisition’s value. Acquiring companies should ensure the AI they buy is not a black box and can be re-created and understood by either their own data scientists or external experts.
Bias, such as gender discrimination in hiring, is one of the biggest risks associated with AI. Bias can show up in data sets that train algorithms either because the data does not represent reality or it reflects existing prejudices. Say AI powers a company’s internal employee recruiting tool, and the algorithm was trained on historical hiring decisions that favored men over women. Chances are the recruiting tool will likely do the same. To reduce such risks, it’s critical for dealmakers to consider plans to tune AI systems to mitigate bias and ensure that decisions adhere to the company’s corporate code of ethics, as well as anti-discrimination regulations. Companies also should continually test for bias in data, models and human use of algorithms.
Traditional controls probably aren’t sufficient to detect the problems that can cause AI bias. Companies should pay particular attention to historical data and data from third parties. Also, data managers should be on the lookout for proxies — biased correlations that can creep into data sets, such as ZIP codes that may correlate to race, or synthetic data that’s created to fill gaps in data sets.
Finally, companies should diversify their teams. Bias is often a matter of perspective, and people from different backgrounds can offer differing views that can help root out bias in the data.
Before considering a transaction, dealmakers need to evaluate the deal's rationale by asking:
If dealmakers decide to buy, they will need to assess the risks and make sure they’re buying what they think they’re buying. This means rethinking the deal process by taking the following steps:
Once the deal closes, the work must continue to preserve the transaction’s optimal value. Start early by developing a comprehensive integration plan that includes the following: