In 2026, AI is already improving individual productivity, but not yet that of most companies. According to McKinsey’s global survey published in August, 80% of respondents see individual productivity gains. However, only 37% can attribute any impact on operating income (EBIT) to AI, the same proportion as a year earlier.

Where is AI being applied to gain productivity?

 

Today, AI performs best in text, data, and repetitive customer service tasks. These are high-volume jobs with clear rules and a human who can quickly review the output.

The most widespread uses in companies are:

  • Writing and summarisation: draft emails, sales proposals, meeting summaries, and long document overviews.
  • Marketing and sales: content pieces, database segmentation, and client call preparation.
  • Customer service: suggested agent responses, ticket classification, and searching the internal knowledge base.
  • Software development: code generation and review, testing, and technical documentation.
  • Analysis and reporting: natural language data queries and creating reports that previously took hours on a spreadsheet.

The change in 2026 is agents: systems that do not just respond, but execute a sequence of steps (search, complete, send, log). McKinsey notes that 40% of large companies are already scaling agents, compared to 22% of smaller ones. Where they report the most cost savings is in supply chain, service operations, and manufacturing.

Why isn’t individual productivity reaching the bottom line?

 

Because saving 30 minutes per person changes nothing if the process remains the same. The freed-up time gets diluted in more meetings, more emails, or simply another unmeasured task.

Macro data shows the same tension. A study by Katz and Jung published in Structural Change and Economic Dynamics concludes that generative AI raises productivity across all worker groups, between 2.5% and 5.7% once adoption reaches critical mass. However, the Federal Reserve Bank of St. Louis notes that total factor productivity in the US grew by just 0.07% over the four quarters ending in March 2026.

The reasonable interpretation: AI works at the task level, and the bottleneck lies within the organisation. McKinsey confirms this from another angle. Only 6% of companies qualify as high performers (more than 5% of EBIT attributable to AI), and that group stands out for redesigning entire workflows instead of adding tools to old processes. Furthermore, 20% already see the costs of operating AI, such as token consumption, as a drag.

How is AI adoption progressing in Australia?

 

It is growing, but unevenly. According to the SME AI Pulse from the National AI Centre, a monthly survey of at least 400 Australian SMEs, 44% were using AI in February 2026. Among those that have already adopted it, broad usage across several areas simultaneously is growing, while single-use adoption is dropping. More complex applications, like agents, remain largely untapped.

 

AI use in Australian SMEs % of adopting companies
Content generation 54%
Data analysis 54%
Cybersecurity and threat detection 48%

The barriers are not about cost. Around 65% of SMEs not using AI distrust its decisions or prefer to maintain human control, and 54% believe it is not relevant to their business. Another 19% simply do not know how to use it. The OECD’s D4SME 2026 survey, covering more than 2,000 SMEs across 12 countries, shows a similar pattern: most use off-the-shelf products, but integration into operations remains uneven.

What do companies seeing results do differently?

 

They select a few processes and redesign them end-to-end. Instead of “giving AI to everyone”, they take a specific workflow, such as processing supplier invoices, and ask which steps disappear, which an agent automates, and where a person must decide.

They also measure from day one. The Federal Reserve Bank of St. Louis analysed around 490,000 earnings calls from US-listed companies and found that nearly 95% of statements regarding AI and productivity speak of future, not achieved, gains. Without a baseline (hours per task, response time, errors), it is impossible to know if the tool pays off.

A typical example is an administration team testing an agent to reconcile supplier invoices. Before starting, they record how many hours a week they dedicate to that task and how many errors occur; during the trial, they measure the same metrics. At the end, they have a concrete number to decide whether to scale, adjust, or discard it.

How to start using AI for productivity in your company?

 

Start with a process, not a tool. This order avoids the most common mistake: paying for licences nobody knows how to use.

  1. Choose a repetitive, measurable process. Something that currently consumes hours, follows clear rules, and has an easily reviewable outcome.
  2. Measure the baseline. How long it takes, how much it costs, and how many errors occur today. Without that number, there is no way to prove improvement.
  3. Organise the data the AI will use. Documents, spreadsheets, and systems must be accessible and up to date. Without organized data, AI has little to work with.
  4. Redesign the workflow with a human in control. Define what the AI does, what a person reviews, and who responds if something goes wrong.
  5. Run a 4 to 8-week test and compare. If the key metric improves, scale it; if not, adjust or discard without guilt.
  6. Train the team and set rules of use. Which data must not be uploaded to external tools, what must always be verified, and how errors are reported.

Frequently Asked Questions

 

Does AI replace employees? For now, evidence does not show this. The Katz and Jung study did not find that generative AI is substituting some groups of workers for others; it is most common for tasks to shift within each role.

How long does it take to see results? A well-measured test shows signals in 4 to 8 weeks. Impact on business results arrives when the redesigned process scales, not before.

Do I need a technical team? For writing and summarisation tools, no. For agents connected to your systems, it is best to have someone who understands data and integration.

Discover how we at Asta can help you design or select your best AI option: https://www.asta.com.au/ai-development

About Our mission in the digital space

 

Asta is a leading full-service technology and consulting agency. We’re trusted industry leaders, who are committed to advancing businesses through powerful IT. Yet, beyond our IT acumen in software, web and mobile app development, our fit-for-purpose managed IT service solutions and our ground-breaking AI and blockchain technologies – there’s something more.

At the core of everything we do is our relentless commitment to people.

Contact and social networks

 

Get in touch with us through our available social channels, and a specialised advisor will contact you to answer all your questions:

Website – Contact – Linkedin – Twitter