By Yemi Olagbaiye, director, client portfolio at digital software consultancy Softwire.
As Seinfeld once said: “a two year old is kind of like having a blender but you don’t have a top for it.” Chances are, a lot of senior execs feel the same about Generative AI (GenAI).
Two years since the launch of ChatGPT, GenAI has been filtering through our everyday work lives at a lightning pace. Yet studies indicate that most business leaders overestimate their knowledge of how GenAI works.
Equally, parenting may look easy from the outside but as any parent trying to manage a toddler tantrum in Tesco or Walmart knows, it’s a lot harder in practice. In both cases, there is a wealth of general advice about raising children and about training AI. However, no bespoke manual is available on how to parent your child or how to manage and train your company’s new purpose-built GenAI solution.
Building on my twin experiences of working with GenAI and being a dad, here are five parenting/AI parallels I’ve uncovered:
Unrealistic expectations v reality: Like the arrival of a new baby, GenAI comes with a lot of anticipation, and it’s easy to get swept up in the excitement. But as the classic Gartner Hype Cycle predicts, you must get past the initial “peak of inflated expectations” before the real work begins. For first-time parents, reality hits with sleepless nights and feeding schedules.
Similarly, on the GenAI front as you start to embed the new technology into your organisation, the initial buzz is rapidly replaced by more mundane considerations, from developing AI policies and uncovering realistic use cases, to freeing up the budget for adapting infrastructure. It’s also important to manage expectations here, whether seeing the Instagram filter of parenthood for what it is (an illusion) or correcting board members on any wild ambitions around what GenAI can achieve.
Sleep training, potty training and nurturing independence: Parenting involves a lot of hands-on training as you help your child to fit into the world around them and become more independent. Similarly, to genuinely adapt and add value to your business, GenAI and Large Learning Models (LLMs) must undergo an intensive and expensive learning process of inputs and responses. This isn’t just about training habits but also teaching independence and the ability to make sound decisions in unpredictable circumstances. As most parents will agree, raising your child is an ongoing process which involves constantly adjusting your approach to fit their capabilities and temperament. Similarly, GenAI models must be continually retrained to adapt to the business itself and the wider business ecosystem.
Ultimately the goal of parenting and of scalable GenAI should be to reduce dependence on constant human oversight. The endgame should be AI systems that can learn and adapt on their own but under well-defined boundaries, similar to how kids learn independence.
Define your values and boundaries: For children to become fulfilled and responsible adults, they must first learn the difference between right and wrong. As parents, this means being clear about your values and boundaries as you teach kids how to interact socially. Similarly, like all rapidly emerging new technologies, GenAI comes with major misbehaviour risks such as hallucinations. These include everything from privacy violations to worker exploitation and the ability to perpetuate existing race and gender biases. Having a clear ethical policy around GenAI and ensuring human intervention and regulation is key to ensuring your GenAI models behave as you want them to.
Expect the unexpected: Whether it’s your five-year-old cutting their younger sister’s hair or your teenager “borrowing” the car, even the best-behaved children can throw parents a curve ball, requiring prompt intervention. Likewise, trained AI models, where you’ve tested or developed a pilot, can behave unpredictably with unexpected consequences. Take Microsoft’s chat AI, Sydney, which admitted to “dark fantasies” and falling in love with users. These “hallucinations” are all the more worrying in the case of black-box AI systems, where users have zero visibility over how decisions are made.
Again, human oversight is key here: anticipating and preparing for unpredictability is as important as responding to it. Parents do this by hiding the scissors or car keys and child-proofing the house. With AI, businesses should plan for AI anomalies by validating, reviewing and refining GenAI outputs and having a robust response framework for when things go wrong.
Remember, it’s a long game: All parents need reminding that their child is a work in progress, on a journey to becoming a fully functioning adult. It’s a journey that involves lots of small steps including mis-steps in pursuit of the end goal. Similarly with GenAI it’s important to have a long term strategy and to start small. There is no point expecting an AI strategy to cover your company’s sales, finance, marketing, legal and software engineering functions from day one. Instead, start by using GenAI applications such as Microsoft Copilot Studio for small tasks. This approach allows your company to develop confidence and expertise in how GenAI models work and then use them to leverage a certain data set, or area of expertise and build your AI capabilities from there.
Next comes trust: When it comes to both kids and AI, building trust is also a gradual process. As children become more capable and reliable, parents start to trust them to walk home from school alone or borrow the car. Likewise, GenAI typically proves reliability over time by delivering consistent results. Businesses need to build trust in their GenAI systems, especially for the stakeholder group’s “sceptics.”
Final thought
It takes a village.
Parenting isn’t done in isolation – as a parent, you rely on family, schools, and communities to help raise your children. AI is similar: businesses should rely on partnerships: consultancies, open-source communities, and collaborations to build robust AI systems that deliver lasting competitive advantage and grow with a company’s evolving needs.
