In a recent study on the future of work, 84% of respondents said they’d be comfortable working alongside AI, and 73% think the word “workforce” should include humans and AI.³ In fact, the study found a large number of companies have already invested in various AI technologies:

• Deep learning - 70%
• Machine learning - 68%
• Support decision making - 67%
• Reaching decisions without humans - 64%

AI can process more information, detect patterns and predict outcomes more accurately than humans. Yet, the human capabilities of understanding contexts and making connections, setting goals, interpreting meaning and crafting strategy are unmatched by AI.

AI is a power tool for developing models and predicting outcomes. Knowing what to do with that information is the domain of humans. Just as humans have to program and train AI, a genuine collaboration requires humans to learn to use AI and keep learning from it continually and, in turn, guide the AI based on outcomes.

A joint study from MIT Sloan Management Review and Boston Consulting Group (BCG) found that organizations that utilize human-AI collaboration stand to gain the most from AI. Only about 10% of the companies that deploy AI see significant financial benefits — between 5% and 10% of their total revenue.⁴ 

That 10% of companies take a different approach to working with AI by learning alongside them.

These organizations:

• Design AI systems that are specific to their company’s needs.
• Build a collaborative way the users can earn from AI.
• Engineer the AI to learn from humans.

The future of work report affirms this result:

• 64% of employees believe they’ll need to learn how to use AI.
• 56% of employees say they’ll need to learn how to train AI.
• 83% of employees say that AI and IT will be collaborative functions.

Companies will have to look at AI beyond automation to derive the most financial benefits. They have to consider how to learn from and transform their organizations with the help of AI.

MIT Sloan Management Review and (BCG) report states that to create a symbiotic relationship, companies cannot just teach AI what humans already know. Human-AI systems have to function based on what a specific situation calls for and adapt to changing contexts.

The study pins down the levels of human and AI collaboration⁵:

• Automator: when AI has all the contextual information, it can decide without human intervention, e.g., dynamic pricing and algorithmic ad displays.
• Decider: when AI has all the contexts, but a human touchpoint is needed, and humans should decide, e.g., call center optimization and predictive maintenance.
• Recommender: when a situation calls for multiple repetitive decisions, but AI does not have all the context, it can recommend, and humans will decide, e.g., sales and operations planning.
• Illuminate: when creative work can benefit from machine learning, and humans use insights generated by AI, e.g., product design by analyzing customer usage data.
• Evaluator: in high-stakes situations when there isn’t enough context, humans can generate scenarios for AI to evaluate, e.g., large seasonal promotions.

Another interesting find from this study is that companies gaining the most from their AI spend 10% of their investment on algorithms, 20% on technologies and 70% on embedding the AI into their business workflow and agile ways of working. That is, successful AI collaboration requires twice as much investment in people and processes than the technology itself.