AI in Business: Key Takeaways from the AWS GenAI Hackathon
We recently took part in the AWS AI Hackathon in Auckland.
We assembled our dream team of Sam Sehnert, CTO, Sid Sankar our Senior DevOps Engineer and Ravindu Thaveesha, a Data Scientist and Developer here at Custom D with a deep understanding, qualifications and practical experience with machine learning.
There were 7 teams competing on the day and mostly from much bigger companies than lil old Custom D, so this was very much a David and Goliath situation.
The goal was to get more familiar with AWS Gen AI services, while solving one of three real-world use cases that were revealed on the day.
The Judging criteria
- Solution overview and value proposition: 25
- End to end thought process: 15
- Working demo (MVP): 20
- Cost and security: 10
- WOW factor bonus points (address multiple barriers): 20
- Scalability of the solution (GTM): 10
Here was our early plan
Business Use Case
Policy Management in Healthcare - Conformity
Problem Statement
Managing internal company policy at scale is a time and labour intensive process. Each time a new policy is added, or an existing policy is amended, the existing company policy needs to be checked to ensure there are no contradictions or double-ups. It also needs to be cross-checked against NZ regulations and legislation for the same reasons.
Outcomes / Impact
A large reduction in manual labour hours and company policies that are up to date and compliant with current NZ regulations and legislation. This reduces internal risk and boosts productivity.
Plan
- Targeting use case number one, we'll generate fictitious internal policy for a company operating in the healthcare space using a large language model (Claude).
- We'll create a knowledge base trained on the generated policies and also NZ regulations/legislation.
- Users of the system will be able to submit policies to cross-check for contradictions or double-ups, and against NZ legislation and regulations.
Confidence rating at 11.30am
Sam: I don't really have much of a sense of what others are doing at this point.
Josh: I heard someone say "Excel spreadsheet" earlier 😀
Pivot alert!
At midday the team changed tack slightly.
The concept was revised so that users could check company statements, communications or other content against internal company policies and NZ legislation.
A real world example of this is scanning a patient letter, analysing customer support transcripts or reviewing training materials before it goes out the door.
This is to reduce the risk of lawsuits, incorrect information and mistakes from happening at scale.
 |  |
Plan B
It was coming down to the wire. With less than an hour left, the team were hitting token limits with the agent instructions. Sid activated plan B and worked on shortening the length to get us over the line.
Turns out it was because we'd been supplied with a free tier account 🤦♂️
Presentations
After tools down, we all took it in turn to present our work.
The presentation criteria was "pretend you are talking to a marketing team or a line of business - someone that doesn't understand what a container or Lambda function is".
Some of the presentations were pretty slick and had obviously done some pre-planning going in, as had we - boys scouts are always prepared!
Sam, our public speaker extraordinaire, hit the stage and did a great job presenting our pitch.
Confidence rating at 4.30pm
Ravindu: We have a chance
And we won!
After a tense wait while the judges deliberated ,the boys took out the gold medal.
 |  |
The judges praised "Conformity" as one of the more mature offerings, highlighting the well-crafted slides that effectively showcased the product. One judge even remarked, “We need this in the market.”
It was a fantastic day, and an incredible opportunity for us to showcase our development skills against the big corporations. We may be small, but we are mighty. 😉
If you’re interested in discovering how AI can be utilised in your business, drop us a message.
