AI Industry & Applications · 2026-01-15 · 05:00

Skybots: from RPA to LLM-powered customer service

Speaker
Skybots
AISG LADP participating company
Type
Industry Leader

In Brief

Accounting-tech firm Skybots used LADP to upgrade RPA into LLM-powered customer service, handling complex accounting workflow queries.

Key Takeaways

  • Skybots is an accounting-tech firm built by accountants; its RPA tools hit a ceiling on tasks needing judgement or variable formats.
  • After learning prompt engineering through LADP, Skybots turned standard chatbots into domain assistants for accounting, tax, audit and ESG.
  • Tasks that used to take two months to code now rebuild in two days; what took two days now finishes in five minutes.

Summary

Skybots is an accounting-technology firm founded by accountants who built RPA to remove manual work. Pure rule-based tasks ran well, but supplier invoices in varied PDF formats broke template logic. Once judgement was required, the bots stalled, and many workflows became semi-manual.

Through AISG's LADP, the team learned prompt engineering and turned generic chatbots into domain assistants grounded in accounting standards, tax regulations, audit rules and ESG frameworks. They published a four-page playbook free for the accounting community, with answers cited back to specific statutes. Code that used to take two months now rebuilds in two days. Workflows that took two days now run in five minutes. Skybots integrated AI directly into its RPA training and now teaches in Hong Kong, Africa and Bhutan. The founder's advice to other firms is blunt: you have to get involved with AI, and the earlier you start the more competitive you become.

Full transcript

Caption language: en · Fetched: 2026-05-02

Skybots is an accounting technology firm combining deep domain expertise with automation. We are accountants who build robotic process automation solutions to eliminate tedious manual tasks. Before LAP, we hit a ceiling with RPA and it handles purely rule-based task well, but it struggles when judgment or variability is involved. For example, supply invoices in PDF forms often have wildly different formats. So simple template rules break down. So once judgment or interpretation was needed, the bots couldn't handle it. So as a result, many workflows became semianual, slowing us down and [music] reducing efficiency. LAP taught us prompt engineering like how to design prompts so that a standard chatbot [music] becomes supercharged. These chat bots act as intelligent assistants for our team.

They can understand [music] complex queries and return precise answers grounded in accounting standards, tax regulations, audit [music] rules, and ESG frameworks. We produce a detailed four-page playbook on how to train, shape, and instruct. We made these tools freely available to the accounting community and thanks [music] to our specialized training, getting more accurate answers than generic chat GPT for domain specific questions. Once PDFs from different vendors and formats can be reliably converted into structured Excel, it removes huge manual effort. And what used to [music] take hours of copying and pasting, formatting, this is now largely automated. [music] And our role has changed from being the preparer to being the checker. We validate that the AI's abstraction [music] is correct, catch anomalies, and fine-tune the logic.

And that change alone frees up significant bandwidth. In accounting, tax, audit, and compliance, precision and justification matter. [music] And our chat bots don't just generate plausible answers. We anchor responses to authorative sources, statutes, standards, and regulations. We [music] invest significant effort in curating and uploading correct documents. That way, when the chatbot answers a question, it [music] can point to the right section of the regulation or standard, giving users confidence in its output. AI has [music] transformed how we build automation. Tasks that used to take me 2 months to code from scratch [music] can now be rebuilt in 2 days using the chat GPT assisted approach. What's very interesting is that BS that took me 2 days to build in the past can be done now in only 5 minutes.

Our development cycle is much faster, so our delivery is much faster, too. Beyond development speed, [music] AI has also revolutionized our RPA training materials. We now integrate AI directly into our RPA courses, showing participants how to combine both technologies for maximum [music] impact. This has elevated our skybox training to the global stage where we have conducted classes in multiple countries. So from Hong Kong to Africa to all the way to Bhutan, users are tapping into our chat bots. The best part is that participants can immediately apply what they learned to solve their real world workflow challenges using both RPA and AI. This project has [music] shifted the DNA of Skyots. Every day we now ask, can AI or automation solve this? After LAP, [music] we can now build AI solutions to eliminate tedious and manual tasks.

We conduct training now to teach non-programmers how to use AI in a practical way to solve many business challenges and many of them are building their own automations using our methodologies learned from LAP. Our experience with the AISG [music] team was positive and supportive. As a non-programmer, I occasionally needed things explained in more detail. But the AISG team was patient [music] and respectful, never dismissive of any question. Thanks to that, I was able to truly grasp the concepts and execute with confidence. My advice to other companies is simple. You have to get involved with AI. You can't [music] keep running a business the same way you did 5 or 10 years ago. AI has totally changed the game. The earlier you start, the faster you will adapt and the more competitive you will be.

Related Videos