Free AI Trials Can Bring Users, but Habit Brings Results
If you run a Malaysian SME, you may have noticed how quickly employees adopt a free AI tool. Someone recommends an assistant for writing proposals, another person uses it to summarise documents, and soon several teams are trying it.
The difficulty starts when the free period ends. Do people still use the tool? Does it support real business work, or was it only popular because access was easy? More importantly, can you tell whether the tool is helping your business, rather than simply adding another app to your workflow?
Perplexity’s partnership with Airtel in India provides a useful example. The company offered a 12-month premium subscription to Airtel customers, and millions of people tried the service. The interesting lesson for you is not the size of the promotion. It is what happened after the promotion: usage remained above its earlier level, while downloads dropped sharply.
TL;DR
Free access can help your team form new habits, but adoption alone does not prove business value. Before introducing an AI trial, define useful tasks, track repeat usage, and prepare a clear decision before access expires.
The strongest test is whether your staff continues using the tool for important, repeatable work after the free period ends.
What This Means
Perplexity partnered with Airtel in July 2025 to provide a free 12-month Perplexity Pro subscription to Airtel’s 360 million customers. New redemptions ended on January 16, while users kept access for 12 months from activation. Source: TechCrunch
The offer produced a significant increase in attention. Perplexity recorded 5.9 million app downloads in India in July 2025, which was 625% higher than the previous month. During the seven months when new users could claim the offer, it recorded an estimated 56 million downloads. Source: TechCrunch
However, downloads are only a first-step measurement. After the offer closed, downloads in India fell by more than 90% during the following six-month period. Yet monthly active users remained nearly 14 million in July, compared with an average of about 2.6 million during the first half of 2025. Source: TechCrunch
That difference matters. Many people may download an app because an offer is available, but only regular users develop a habit. For an SME, the equivalent is an employee trying an AI writing assistant once compared with using it every week to prepare customer replies, internal reports, or sales material.
Free access creates an opportunity to test a habit. It does not automatically create a useful business process.
How This Applies to Malaysian SMEs
First, use trials to solve a specific operational problem. If your sales team spends too much time preparing quotations, a trial should focus on drafting quotation explanations, organising customer requirements, and preparing follow-up messages. If your administration team handles repeated questions, test the tool on standard replies and internal procedures. Avoid giving everyone access without direction. That usually produces scattered experiments that are difficult to evaluate.
Second, consider the language needs of your customers. Malaysian SMEs often communicate in English, Bahasa Malaysia, Mandarin, Tamil, or a mixture of languages depending on the customer and industry. Test whether the AI tool produces suitable wording for your audience, not merely grammatically correct sentences. A response that sounds too formal, too direct, or culturally unsuitable can create extra editing work. Ask staff to compare the original draft, the AI-assisted version, and the final message sent to the customer.
Third, measure repeated use by role. A small wholesale company might have one person using AI for product descriptions, another for stock-related customer replies, and the owner using it for meeting summaries. These are different use cases. Track which tasks are repeated each week and whether staff complete them faster or with fewer revisions. You do not need a complex analytics system. A shared spreadsheet with the employee, task, frequency, and result is enough for an initial trial.
Fourth, set rules for confidential information. Do not allow employees to paste customer identity card numbers, bank details, passwords, private contracts, or sensitive staff information into an AI service without checking your company’s policy and the provider’s handling terms. Start with low-risk material such as generic marketing drafts, public product information, meeting agendas, and anonymised examples. This lets you test usefulness without exposing information your business must protect.
Finally, plan the end of the trial before it begins. The Airtel example shows why the expiry date matters. Some users may continue because they genuinely value the service, while others may be renewed automatically or may not realise access has changed. In your business, schedule a review before the trial ends. Ask each user to demonstrate two or three completed tasks and explain what would happen if the tool disappeared tomorrow.
A Simple Measurement Framework
Use the following structure to evaluate an AI trial. The figures below come from the Perplexity case and are included as reference points; your own targets should reflect your workflow and team size. Source: TechCrunch
| Measure | Reported Perplexity result | SME question |
|---|---|---|
| Initial downloads | 5.9 million in one month | How many employees actually started the trial? |
| Monthly active users | About 22 million at its peak | How many users return for work each week? |
| Later monthly active users | Nearly 14 million | Did usage remain after the promotional peak? |
| Earlier monthly users | About 2.6 million average | Is usage now higher than before the trial? |
| Download decline after the offer | More than 90% | Are new sign-ups less important than existing user activity? |
These figures are not a direct forecast for your company. They illustrate an important distinction: acquisition tells you how many people started, while retention tells you whether the product became part of their routine.
Practical Takeaways
- Choose one business problem first: select a task that happens regularly and has a clear owner.
- Define success before testing: decide whether you are measuring faster completion, better consistency, fewer errors, or improved response quality.
- Start with a small group: include people who perform the task frequently and can provide practical feedback.
- Record actual examples: save anonymised before-and-after work so you can compare results.
- Check language quality: review Bahasa Malaysia, English, Mandarin, or other customer-facing outputs before sending them.
- Protect sensitive data: create a simple list of information employees must never enter into an AI tool.
- Review usage before expiry: identify active users, occasional users, and people who found no useful application.
- Keep human approval: AI can draft, summarise, and organise, but a responsible employee should check important outputs.
- Document the workflow: write down the approved prompts, review steps, and storage location for final work.
The Bigger Picture
The Perplexity example suggests that a large free distribution campaign can leave a service with a bigger active audience even after the promotion finishes. Its estimated Indian in-app purchase and subscription revenue between February and mid-August was about 60% higher than during the period when new users could claim the offer, although the data could not confirm how many former Airtel users became paying customers. Source: TechCrunch
For Malaysian SMEs, the long-term lesson is to judge software by the workflow it creates, not by the number of people who install it. A free trial is useful when it helps you discover a repeatable process that your team understands and can supervise. It is less useful when employees experiment privately, produce inconsistent results, or stop using the tool as soon as the promotion ends.
AI services will likely continue using free access, bundled subscriptions, and distribution partnerships to attract users. You do not need to follow every offer. Instead, create a simple internal standard: identify the task, test the output, protect business information, measure repeated use, and review the decision before the trial expires.
If the tool helps your team complete important work more consistently, it deserves serious consideration. If it only creates occasional novelty, you have still learned something valuable—provided you measured the trial clearly.
Ready to Streamline Your Operations?
Your business should run itself. AutoRunBiz deploys AI agents to automate your daily operations — WhatsApp orders, invoicing, customer follow-ups, and accounting. Book a free 15-min ops audit to see where automation fits your business →