Almost every organisation, large or small, is either experimenting or using AI in some way. And while there are many pilot projects and proofs of concept testing AI and its potential, there is a significant leap between those trials and scaling to production.
Successful adopters redesign the process instead of attaching AI to the existing one. Max Wessel, the SVP of Product at Workday said, "The biggest difference between companies that get AI right and companies that dabble is companies' willingness to reinvent processes … versus just bolting it on."
As an example, Wessel described a process that can be time consuming and frustrating for HR professionals and job candidates: booking an interview time. Once the interviewer or interviewers have identified a suitable time, which is done automatically by Workday's software, the software sends a text message to the candidate. They can then respond, choosing a suitable time from those offered.
"We schedule 10% of the job interviews in the United States today. And it's via an SMS system," said Wessel.
For finance teams, audits are usually carried out by extracting a sample set of transactions and assessing them against regulatory criteria. This task is repeated and can consume significant time and resources. Wessel said Workday's financial audit agent conducts continuous control testing across all transactions rather than retrospective sampling.
"It runs continuously in the background and enables organisations to move from a retroactive sampling and evidence-collecting perspective to a continuous control-testing model. It's a game changer for how they approach the problem," he said.
Reinventing processes requires people to adapt. And this is a significant issue.
"Humans are not great at change. Getting AI right is around changing how we've been doing things for a very long time," said Wessel.
A key to succeeding with AI is converting domain expertise into agents that can execute entire processes with human oversight. While users in many organisations are creating smaller personal agents, Wessel noted that organisational agents have access to context-rich systems and use multiple skills, history and governance controls.
As well as redesigning business processes to leverage AI, there will be changes in how organisations arrange their personnel and how they evaluate costs and value.
As processes become automated, the way we measure cost and value needs to change. Traditionally, we have paid people at a rate that is tied to how much time they spend on an activity. With AI, once an agent is created, value will be measured on the usefulness of the output. But that can result in people having free time. Wessel said this is where Workday's newly announced Work Intelligence function becomes important.
Work Intelligence can look across systems where work is recorded and give leaders data, so they can redesign roles and support people as their jobs change.
That shift towards outcomes-based measurement rather than time is perhaps where the most fear for people comes. If they can no longer bill or receive a salary for the time it takes to deliver an outcome, they may feel their career is in jeopardy. However, Wessel doesn't think this is happening as quickly as some people anticipate.
"There will be a large shift towards outcomes, but probably over a meaningful time period because it's very easy to measure inputs. Right now, it's very hard to measure outcomes," he said.
For many organisations, shadow AI looms large. And this is causing many organisations to try and apply the brakes to AI. Some have not reviewed policies since the first GPT tools were made public and still retain zero-tolerance for AI.
Wessel's said that as unsanctioned tools continue to enter workplaces, vendors must make trusted systems compelling enough to use. He highlighted the need for non-human identities to be properly managed and guardrails to be put in place.
The challenges this poses is a strong motivator for Workday's acquisition of C1.ai, which was announced at the company's Workday Rising event.
With AI evolving rapidly, models that were considered frontier and bleeding edge just a month ago are already superseded by faster and more accurate models. The coming year will test whether enterprises can turn capability into repeatable process transformation at scale. For that to happen, they will need to ensure they have guardrails in place to ensure agents run in a controlled and reliable way and invest in supporting people to not just use the technology well but to ensure their domain expertise is not lost and is used in other parts of the organisation where they can add value.
"I think that we're relearning how to organise and how to communicate value," said Wessel. "I think we will have much more dynamic organisational structures and more diffuse skills across what were historically very narrow job boundaries. And that's part of the excitement of this moment."