5 lessons in AI productivity
Many organisations are on a journey to use generative AI (GenAI) to unlock greater efficiency and productivity. In fact, 81% of organisations around the globe plan to increase their investments in AI over the next year, according to the Enterprise AI Maturity Index by ServiceNow and Oxford Economics. But how?
Consider our experience at ServiceNow. In just a few months, we achieved more than $10 million in tangible benefits and enhanced productivity by using GenAI. The technology is now performing work equivalent to 50 full-time employees on an annualised basis, reducing manual tasks and creating more time for our teams to focus on higher-value work. Let’s look at some of our lessons from the pursuit of AI productivity.
1. Repetitive tasks are ripe for AI
Every job is, at its heart, made up of a series of tasks. If you’re a chef, you plan a menu, source ingredients, prep food, cook dishes and plate and serve them to customers. If you’re a salesperson, you typically start by researching potential clients, then create a compelling pitch, follow up with a tailored demo for the customer, and finally negotiate pricing and close the deal.
Some of these tasks are clearly ripe for automation and technology augmentation. The question is: Which tasks are most suitable for people and which should be delegated to machines?
Tasks that fall into the latter category are repetitive in nature, heavily dependent on data, and involve synthesis, predictions, or recommendation. These can be handled by AI with some acceptable margin of error.
For example, personalised follow-ups to customers make up a significant portion of a salesperson’s job. At ServiceNow, we deployed GenAI for this outreach and saw a twofold increase in the response rate.
Since then, we started using GenAI for digital personalisation. One example is our use of a capability we call Email Assist, which tailors messages from sales teams to better resonate with customers based on a range of traits, including their industry, job role, and individual behaviours.
As a result, our sales teams can spend more time on core tasks: developing new business and maintaining productive customer relationships. In fact, we estimate this tool has helped us reduce time spent per sales lead by 60%.
2. Simplifying search has great value
Across virtually all knowledge worker roles, there’s one task that’s especially ripe for machine augmentation: searching for information.
The unpleasant reality for many knowledge workers is that finding information is a frustrating scavenger hunt. We all spend a substantial amount of time searching for data and content.
We faced this same challenge at ServiceNow. When we deployed GenAI to streamline our approach to search, employees gained a powerful tool to easily find answers on their own. This has saved time for colleagues who previously had to answer those queries. In short, we’ve seen a productivity boost for both the requester and the fulfiller of information.
3. Process mining can speed results
Over the past 25 years, digital transformation has led to extensive automation of countless processes, generating vast amounts of data along the way. Many organisations have essentially hoarded this data like squirrels hoarding nuts. Process mining can help make sense of this data, identifying inefficiencies and ways to enhance speed.
When we implemented process mining and optimisation at ServiceNow, the technology performed well, but we initially observed minimal impact. We didn’t have a clear process owner in place or a way to organise the next steps after inefficiencies were found.
To fix this, we developed a continuous improvement practice through which the process owner declares the “so what”—the specific actions needed to remove the bottleneck. This approach has led to a more than 20% improvement in process speed in just a few months.
4. AI should be integrated with experience
A key component to driving AI productivity is integration with the user experience. Cloud-based software as a service (SaaS) platforms have rolled out customisable out-of-the-box AI models integrated into their user experiences.
In the past, custom-built AI models allowed organisations to tailor solutions for specific business needs, using unique data sets and algorithms. Creating these models has traditionally taken weeks. Now SaaS platforms offer out-of-the-box models that can be deployed in days or hours.
These platforms eliminate the need for extensive coding and facilitate end-to-end use cases. Using the ServiceNow platform, we’ve achieved a remarkable 70% reduction in model development time.
5. AI measurement is crucial
There’s no shortage of noise around AI in the marketplace today. What separates the hype from reality? Maths. Rigorous measurement is crucial.
Our view at ServiceNow is that we must manage our AI models the same way we manage human talent. In other words, we need to hold the models accountable for performance. For instance, we use an AI governance app to track model approvals, performance, and value, allowing us to continue using effective models and retire those that underperform.
Deploying GenAI at ServiceNow has played a key role in yielding millions of dollars in just a few months, driven by improved search, virtual assistants, and developer tools.
As organisations embrace AI, those that focus on accelerating their productivity stand to reap enormous benefits. At ServiceNow, we anticipate that AI will augment our capabilities and allow us to focus on work that’s more strategic, creative, and even fun.
Find out more about how ServiceNow helps organisations put AI to work.