Many documentation teams experiment with AI to draft content, summarise information, or refine wording—but occasional AI use rarely delivers meaningful productivity gains.
In this expert opinion, we spoke with a member of the ELEKS Information Development Office about what it really takes to see ROI from AI in documentation. Based on insights from a recent proof of concept, they explain how to maximise the value of AI in technical communication.
With over 15 years of experience in technical communication and content strategy, Lesia specialises in user-centred documentation and promotes AI-assisted documentation processes that improve efficiency while maintaining high editorial standards.
How widely is AI used in technical communication teams today?
Lesia Zasadna: If we’re looking at recent surveys from communities like the Content Wrangler or Write the Docs, roughly 70–75% of technical writers say they’ve experimented with generative AI tools, mostly for drafting content, summarising information, or refining wording.
But those surveys also suggest that for many people, it’s still something they reach for occasionally. And when AI is used that way, a lot of its potential simply goes untapped.
So what needs to change for AI to actually deliver measurable ROI in documentation?
LZ: If someone opens a chatbot to help draft a paragraph, summarise a document, or rewrite a section, that can save a few minutes here and there.
But to see real ROI, AI has to become part of the workflow itself. That means looking at where the biggest time sinks are and then deliberately integrating AI into those stages.
When you start breaking documentation work down that way—identifying the areas that consume more time than you like, setting clear time limits for them, and then figuring out how AI can help stay within those limits—that’s when things start to move forward. At least that was the idea behind our proof of concept.
What exactly did you test during your PoC, and what did you find?
LZ: Our idea was simple: if we want to minimise our overall documentation effort, we need to apply AI to the most time-consuming documentation activities:
- Analysing source materials to grasp the core data and filter out any product-centric or process details that the user won’t be interested in.
- Structuring the remaining information into a logical outline and identifying any gaps.
- Drafting the content, one section at a time, from the user-centric perspective.
We saw the biggest productivity gain during the analysis phase. In our experiment, this stage became roughly 2–3 times faster when using an AI document generator.
In the structuring stage, AI was helpful too: we managed to create a solid outline about 1.5–2 times faster. But this work is a relatively small, one-time effort, so the time saved there didn’t change the big picture dramatically.
The most interesting stage was content drafting. For the sake of the experiment, we decided we wouldn’t write a single word of the guide ourselves. Instead, we acted as curators of the process—designing prompts and refining them until AI produced output close to what an experienced Information Developer would write.
The thing is, AI is not naturally wired to write the way we do. In good documentation, minimalism is essential, so every single word has to literally earn its place. The information should also be presented in a way that allows the readers to grasp the core message even when distractedly skimming. That’s why getting to that level of clarity required lots of back-and-forth prompting. Even so, the drafting stage still ended up about 1.2–1.7 times faster than our fully human process.
So what does it amount to in terms of overall documentation effort?
LZ: Well, that depends a bit on the type of documentation we’re talking about. Tutorials, help centres, and AI-assisted chatbots – they all have slightly different workflows.
If we take something fairly typical – a user guide – and the total documentation effort is 100%, where does the time actually go? In our experience, analysis takes about 30–40% of the effort, drafting roughly 40–50%, while outlining and publishing make up the rest.
Which means that with the analysis/structuring/drafting productivity gains we discussed, the AI document generator can save 20–25% of the overall effort.
Or, if we look at documentation work in terms of continuous delivery, an AI document generator can save 2–2.5 working days during every 2-week iteration. The impact grows over time. When the same documentation is updated across multiple release cycles, those time savings accumulate with every version update. And that’s just one document—the larger the documentation set, the greater the gain.
20–25% is more modest than some of the numbers people associate with AI. Why isn’t the savings larger?
LZ: A big part of the answer is that source documentation rarely captures the full context of how users actually interact with the product. A good document not just captures the procedure steps inside the product but places them in the context of what the user is trying to achieve and what obstacles they may come across in the process. A lot of that knowledge lives in people’s heads rather than in written materials. So you still need to talk to stakeholders and fill in those gaps.
Verification is another reason. Even if AI helps analyse materials and draft content, you still have to check everything against the actual product. If you’re using AI responsibly, you still need to confirm that the documentation reflects the real behaviour of the system and doesn’t subtly invent features that don’t exist.
And then there are tasks that remain largely manual: capturing screenshots, shaping the visual layout of documentation, configuring publishing pipelines, and so on. Some of these steps can be partially automated, but in most real projects, they still require a lot of manual work.
So if we’re talking about high-quality documentation, the 20–25% gain we measured is a careful but realistic estimate.
That said, I do think the potential can grow over time. Integrating AI even deeper into the documentation workflows and using AI agents expands the opportunities to automate documentation work even more.
What advice would you give organisations that want to see real ROI from AI document generators?
LZ: Start with the workflow: look at where your documentation team actually spends the most time, and integrate AI deliberately into those stages instead of using it occasionally.
Just as importantly, keep humans responsible for verification and quality. AI is excellent at accelerating analysis and drafting, but it works best when experienced writers guide the process, measure the results over time, don’t despair after a few not-good-enough results and really dig their heels into maximising AI usage in documentation.
Related insights
The breadth of knowledge and understanding that ELEKS has within its walls allows us to leverage that expertise to make superior deliverables for our customers. When you work with ELEKS, you are working with the top 1% of the aptitude and engineering excellence of the whole country.
Right from the start, we really liked ELEKS’ commitment and engagement. They came to us with their best people to try to understand our context, our business idea, and developed the first prototype with us. They were very professional and very customer oriented. I think, without ELEKS it probably would not have been possible to have such a successful product in such a short period of time.
ELEKS has been involved in the development of a number of our consumer-facing websites and mobile applications that allow our customers to easily track their shipments, get the information they need as well as stay in touch with us. We’ve appreciated the level of ELEKS’ expertise, responsiveness and attention to details.
