Deploying Human-Centric AI in the Workplace
It may well be that business is approaching the trough of disillusionment that technology research firm Gartner has often be quoted for. Even Microsoft CEO Satya Nadella has started to downplay it's AI hyperbole, pumping the breaks on AI at work. It's not the technology itself, although sometimes it is, I think has more to do with how businesses, across multiple industries, haven't figure out how to integrate AI into the workplace.
It hasn't helped when Venture Capitalists and Fortune 100 CEO's gleefully declare to shareholders that they're about to sack huge swathes of their workforce. That may delight shareholders, but it's not really an employee motivational approach.
When we understand that technologies are just tools to aid in human problem solving, rather than a solution itself, this can fundamentally shift how a business approaches digital transformations and how to deploy AI in the workplace.
This enables a business to develop a framework that focuses on human capabilities, workplace cultural impacts and longer term benefits to the company. It also improves the odds of a more successful deployment.
A Framework for Human-Centric AI Deployment
Consider the following as you look at how to bring AI into the workplace more successfully. If you approach it as a job replacement tool, given most AI capabilities today, you'll end up with a failure. Very few instances of AI can replace jobs. Perhaps some, but mostly in low-skilled areas.
There's enough research now that shows lower performing workers will find improvements when AI is used to assist and augment their work. Interestingly, for high performing workers, AI doesn't help much.
Cultural Integration: Start by looking at and understanding workplace culture. The different networks in the organisation and actual power dynamics, not the often unrealistic organisational chart. Recognise that work practices are part of a person's identity.
Enhancement Mechanisms: Find areas of high-friction in the company. AI tools can help reduce friction in systems and reduce cognitive load. Look at feedback loops where AI works with humans in systems.
Reciprocity: All organisations are systems and in all organisations are systems of recirptocity. Like when different departments help each other, or employees. AI systems and agents can become part of this reciprocity system. AI and humans can learn and help each other. Find those opportunities.
Decision Making: For years many companies have lived under the naive illusion that more data is better. That is rarely the case. Ai tools hallucinate, especially with more data. Humans are often going to make better high-risk decisions than AI tools. Remember, technology doesn't solve problems, humans do. Use AI tools for bureaucratic, boring, process driven activities that humans don't like.
Support not Replace: Your organisation relies more heavily on social bonds that facilitate knowledge transfer, not matter how good your SharePoint or knowledge management system is. AI tools should support knowledge management and transfer, not replace it. AI systems don't create new knowledge, they just parrot what we already know. Humans create and share new knowledge.
Enable Modification: No AI tool, like any technology like software, will resolve all problems. If you use an ERP system like SAP or Oracle, then you know this. Enable teams to modify the AI tools they're using. You may be surprised how much better both the team and AI system perform.
While I do have a formalised framework for bringing AI into the organisation, this is just to cover some high-level thinking you need to take. Beware IT firms and consultancies that claim massive success rates at implementing AI into the workplace. Generative AI, like ChatGPT, Claude etc., have only been around just over two years. That's simply not enough time to provide empirical evidence of success by anyone. It's marketing bullshit. Don't fall for it.
The reality is, everyone is still trying to figure it out. Models, Frameworks and systems need to be malleable, adaptive and evolving. Digital transformation methodologies for software and enterprise systems will not work for implementing AI in the business. That much I do know.