This is the second article in our 10-part series on the key questions posed by CHROs and senior HR leaders about AI implementation (see bottom of the article for links to the others). The series explores the implications for work, talent, trust, ethics, regulation and human connection. Each article stands alone, and together they build a practical view of what responsible, commercially useful AI adoption requires. In the interests of transparency and openness the data sources for all ten articles are provided at the end of each article.
Key points:
- The goal is higher levels of organisational performance through the human-machine collaboration, not simply automation.
- AI should enable “superworkers”: people able to work faster, smarter and across disciplines.
- Software alone is not enough; to get the full potential of AI, the operating model will need to be redesigned around AI-enabled work.
- Rigid hierarchies and silos will blunt AI’s impact because they slow the speed of decisions and lack the necessary collaboration on which successful deployment of AI is based.
- Decision-making needs to move closer to the point of action, where people can act quickly with AI support.
- Repetitive work should be centralised and simplified, freeing people for higher-value activity.
- Fusion teams will become critical, blending business expertise, technology skills and agile problem-solving.
- HR’s role is to map talent, skills and structure to this new AI-enabled way of working.
The ultimate strategic objective of AI enterprise implementation is not merely to automate repetitive administrative tasks, but to unlock a profound state of organisational "Superagency". Rather than see AI as a negative threat that so many people do, Reid Hoffman proposes a much more positive take and suggests that AI provides individuals with "superpowers," allowing them to achieve goals more effectively and, in turn, creating a more prosperous society. In this perspective, the workforce transitions to become "superworkers", individuals capable of executing highly complex, cross-disciplinary workflows at speeds previously considered impossible. This depends on people having the capability to deploy AI effectively, and the organisation integrating AI in a way that makes accessibility and usability as seamless and easy as possible. It starts not with AI but on what you would like people to be able to do and how integrating AI can support this. This is substantially more than the mere procurement of software licenses; it demands a complete deconstruction and reinvention of the corporate operating model. This requires the type of integration we discussed in the first article. It also requires a fundamental transformation of the organisation (about which we have written here).
Traditional, rigid hierarchical structures are commonly barriers to the successful adoption of technological innovation. Each of these barriers is a ‘human-created’ handover point setting up the need for communication, a work flow and human intervention. AI operates at best in a world with limited/no boundaries. There is of course some necessity to these boundaries. Current processes tend to rely heavily on compartmentalised data (functional scorecards), reinforce departmental silos, and restrict the lateral, high-speed flow of information. This effectively neutralises the cross-functional capabilities of generative AI models. Generative AI models require significantly flatter, less rigid organisational hierarchies that aggressively decentralise decision-making authority, pushing autonomy to the very edges of the workforce. Instead of requiring managers to approve every action, use AI to set operational parameters. Employees have total autonomy to make decisions within those guardrails, only escalating to leadership when an exception occurs. A failure to do this is a failure to liberate the full potential of AI. For decision-making authority to be delegated effectively in this way means people need to have the skills to take advantage of AI. Already many organisations are failing as people ‘play with AI’ uploading sensitive information to freely available platforms in order to help them do their work without realising the ethical and legal implications of so doing. Once the information has been uploaded, all control of that data has been lost. To encourage people to experiment without compromising the organization, create a simple, jargon-free AI Acceptable Use Policy. Make it clear what data is strictly off-limits (e.g., customer data, unreleased financials) and what is safe to use. Creating your own unique and protected private AI or ‘enterprise AI sandbox‘ (as they are sometimes known) solves this problem.
At the same time as decentralisation of decision making there needs to be a corresponding centralisation of core, repetitive corporate functions to ruthlessly eliminate duplication of administrative work, thereby freeing key people to focus entirely on higher-level initiatives. To this end, we are also seeing the emergence of agile "fusion teams." These are highly multidisciplinary groups that seamlessly combine deep domain expertise with advanced technological proficiency to iteratively solve complex business challenges in real-time. Such groups of people will not ‘magically’ work together without some help and development. From our experiences, bringing together people with different backgrounds (e.g. finance, IT, HR, engineering etc can be problematic due to the different ways these disciplines approach resolving challenges. We find it useful to run "Ways of Working" workshops with multidisciplinary teams. It’s useful to mandate a session to define shared language, establish conflict resolution methods, and align on goals. It’s also useful to reward the collective output of the fusion team rather than individual departmental metrics.
From an HR perspective, the challenge is mapping these radically new ways of operating into structural designs, and mapping these to existing talent pools, ensuring that the organisation possesses the requisite agility and cross-functional literacy to take advantage of AI. All whilst helping individual leaders and teams ‘make sense’ of their work, their accountabilities. For example, break down jobs into a series of tasks, identifying which tasks AI can automate, which it can augment, and which require distinct human judgment. It’s also helpful to find the employees who are naturally excelling at human-machine collaboration as they can become peer coaches, helping upskill those who are struggling or resistant to AI. This type of resistance is nothing new, many years ago the advent of IT systems in organisations that hitherto were paper based faced similar ‘organ rejection’ from the use of technology.
If AI centralizes and automates routine administrative tasks, HR will need to figure out how junior employees will learn the ropes. For example, create mentorship models that focus on critical thinking and AI oversight rather than data entry.
Questions you may want to think about:
- How prepared is our current leadership culture to shift from a rigid, manager-approval model to one where decision-making authority is pushed to the edge, governed by clear AI operational guardrails?
- To what extent does organising work around product groups, value chains, information flows and or programmes aid or counter the current AI-led design rethink?
- What’s the role of the current governance structures and resources in steering not throttling an organisation's AI ambition?
- How do you retain meaning in role accountabilities that may increasingly become horizontally/workflow aligned (rather than functionally led)?
- If AI is as transformative organisationally as suggested? Who needs to be around the table when designing the organisation of the future? What skills do they need ahead of this conversation?
- Do we currently have a simple, jargon-free AI Acceptable Use Policy, or is our workforce operating in a shadow-IT environment—uploading sensitive corporate data to public platforms just to keep up with their workloads?
- As AI centralizes and automates the routine administrative tasks traditionally used to train entry-level employees, how must we redesign our junior roles and mentorship models to ensure the next generation still develops foundational expertise and critical thinking?