For a while now, it is not easy to keep the conversation about AI and work at a safe distance. As long as it was about productivity gains, new tools or future skills it looked far and abstract enough to not bother us much. But there are a growing number of discussions I have these weeks when the topic is related more to the org chart. Companies are starting to change how they hire, where they invest, and which roles they expect AI to absorb. Some jobs are being cut, and some work is being rearranged.
But before deciding what happens to people, we had better understand what is happening to the work itself. I am aware of some examples from several large companies that are already showing different versions of this shift toward a new way of thinking about AI implementations. Here’s a few:
Klarna is one of the clearest examples. Its AI assistant handled 2.3 million customer-service conversations in one month, covering two-thirds of customer-service chats and doing work equivalent to 700 full-time agents. That number is impressive but has also raised some newfound leadership questions: which customer interactions can be automated, and which moments still need a person because trust, frustration, context, or sensitivity are involved?
Amazon shows another side of the story. Generative AI and agents are starting to handle corporate routines like reporting, coordination, analysis and documentation. Andy Jassy has described a future where fewer people are needed in some roles, and more people are needed in others which means that AI also changes where work should sit.
Shopify flipped its hiring logic by requesting teams to show why AI cannot help to get the work done before asking for more headcount. So that changes the order of the conversation. Hiring is no longer the automatic next step when workload grows but teams should first ask if the work has already been redesigned or not?
IBM has paused hiring for roles where jobs could eventually be replaced by AI and automation, especially in back-office functions. This kind of workforce change begins with a vacancy that is not refilled, a team that does not expand, or a role that slowly disappears from the future organization chart.
These four examples all point in the same direction: AI-driven organizational design is both a technology question and a leadership / HR / finance / operations, i.e. a broad boardroom question.
To help leaders map their possibilities, I have created a model, easy to remember when you want to reshape your organization, just turn to the SHAPE model.
S – Stop or slow hiring
Some roles should no longer grow automatically.
If a person leaves, the old reflex is to refill the same position but AI challenges that reflex. We should explore first if the work has changed enough to redesign it?
This is the lesson from IBM’s example. Sometimes AI restructuring starts through hiring restraints before it becomes visible as formal restructuring.
H – Hold human work
Some work should stay strongly human.
This includes work that depends on judgment, trust, empathy, ethics, accountability, and client relationships. It also includes moments where the situation is sensitive, ambiguous, emotionally loaded, or commercially important.
Klarna’s example is useful here because it shows both sides. AI can carry enormous volume but at the same time, the human layer protects the customer experience and the brand. The point is to know where human presence creates real value.
A – Automate
Some work is clearly ready for automation.
Repetitive, rules-based, high-volume tasks are the natural candidates. Customer-service triage, standard documentation, first-draft reporting, simple internal requests, data summaries, routine analysis, and administrative workflows can often be done faster with AI.
Also, automation should remove friction so leaders also need to ask whether automation is reliable enough, safe enough, and good enough from the customer’s or employee’s point of view. It should not create hidden frustration.
P – Pivot resources
AI changes where people, money, and management attention should go.
Some capacity may move away from routine corporate tasks while more capacity may be needed in AI infrastructure, implementation, governance, product development, data quality, cybersecurity, change management, and higher-value client work.
Just like at Amazon: fewer people may be needed in some areas, while new roles and capabilities become more important elsewhere. The workforce moves around.
E – Embed AI
The deepest change comes when AI becomes part of how the company works.
This is what makes the Shopify example interesting. When leaders seek for tasks where AI is already expected to be part of the work, it is no longer treated as a productivity experiment but AI becomes part of the way teams ask for resources, design workflows and define what good work looks like.
The real unit of analysis is the task
I never liked hiding behind job titles but with this new era of AI it seems to me that a job title has become too large a unit of analysis.
A customer-service role, HR role, analyst role, finance role, sales role, or manager role contains many different tasks. Some of them are repetitive, some require context, some carry risk, some create trust precisely because a human being is present. So when thinking about redesigning a job, we should first explore what kind of work make up the role?
After we have that list, leaders can decide what to stop hiring for, what to hold as human, what to automate, where to pivot resources, and how to embed AI into the operating model.
Questions for the executive team
If you are sitting in a board being part of endless conversations about the different option to introduce AI into your organization, let me suggest you a line of questions for a practical discussion:
- Where are we still hiring automatically, even though the work itself may already be changing?
- Which tasks, roles, or customer moments should remain clearly human because trust, judgment, or accountability matter?
- Which repetitive or high-volume tasks could AI handle safely without lowering the quality of the experience?
- Where should we move people, budget, and attention as AI changes the work?
- How do we make AI part of normal management practice, instead of treating it as a separate experiment?
For the past few years AI is already inside the workforce conversation whether we like it or not. Whatever we do, work will be reshaped by AI anyway, so the real question is if we do it it deliberately or rely on the accidental shifts. Speaking for myself, I prefer to drive change.
