AI & Automation - what i've learnt and what i've built (so far)

About 7 months ago I made the decision to take up a new challenge and move on from my Head of Marketing role to join Arkana Energy, one of the top 5 solar installation companies in Australia.

My role there? Use AI, Automation and whatever other tools needed to help improve the business – from marketing, sales, operations – any part of the business that I could help move the needle and make a positive impact on. Plus of course, taking on the performance marketing and lead generation responsibilities and seeing how we can evolve that and the 5,000-7,000 enquiries we generate each month.

So far, it’s been a whirlwind 6-months and along with training the team on AI and how to use it, getting my head across how it really works and can be used in a business (plus it’s shortfalls and where it can’t be yet) plus many many hours spent learning the solar industry in and out, trying to get python code working, fixing prompts and a lot of trial & error ive actually managed to build some things that are in use today – and i’m excited for the next 6-months!

A few things ive learnt

Some of my learnings to date have been (in no particular order):

  • You don’t need to be a developer to develop tools with ChatGPT there to help – however it can take a lot longer.
  • Using the right low-code or no-code software/tools can speed up and help with development when you don’t have a developer.
  • Zapier rocks – it’s my number one tool I use day in and day out – and im ready to join the fan club if it ever launches.
  • Before diving into building a tool or automating something – check you can get it working manually, ie. perfect your prompts and the outputs you are getting in ChatGPT or similar first, prior to moving to the build stage.
  • It’s best to start with one, simple project/application which is solving one, specific and simple business problem. You’ll learn a lot.
  • Large Language Models (ie. ChatGPT) can be very inconsistent – and it’s tough to get a project from working 60-80% well to 95%.
  • With 100s of “AI” tools launching every day (some just sites sitting in front of OpenAI’s API) it can be confusing, overwhelming and hard to know where to start. With so much noise in the industry it’s best to analyse the data and find the biggest business problems and THEN find ways of solving it, not the other way round.
  • Projects never really end and keep requiring tweaking, optimisation and improvements and there are always ways you can evolve them. It will be easy to get stuck on maintaining existing tools and projects vs evolving and moving to new ones.
  • AI input is a great way to improve your existing automation workflows – adding additional data or decision making abilities to help improve what your automation does.
  • I have a lot to learn & can’t do or learn it all – using third party tools, consultants, developers will be essential in the future.

Some things i’ve built

Now onto a few of the projects i’ve built so far (without an AI team or developer!) which may give you some inspiration for things to build for your own business.

AI Call Transcription, Analysis & Call Centre Coaching

Automated AI call analysis and reporting to assist with training and ongoing coaching of call centre staff – five different ways to analyse a call including comparing against script, against a good/bad call checklist, and custom questions/analysis.

AI Driven Inbound Email Processing & Automation

Automating the processing of inbound emails to generic/company email addresses – with AI categorising and deciding next best steps and pushing through Zapier automation paths with majority of inbound emails completely automated.

AI Outbound 1-1 SMS Lead Gen Campaigns

Through a partner solution we’ve launched an AI tool that sends 1-to-1 outbound SMS campaigns to reactivate old and uncontactable leads including objection handling and upselling add-ons (ie. a battery) where a user has already bought. Automatic call bookings and CRM updating on successful conversation.

AI Built & Written Customer Support Portal

New support portal for customers built and written by AI to help prospects and customers get pre-purchase info, and guides on making the most of their purchase + troubleshooting. Now also used for training AI on company approved data for other projects. 125+ articles and counting.

AI Support Ticket Processing + Initial Customer Replies (coming soon)

Support ticket processing and ticket enhancement (adding historical info about products used in install, install date, installer notes etc). Initial customer replies by AI to request missing info and push towards relevant support portal articles to self-solve before ticket can be reviewed by a human – are coming soon (being sent to just me for now until we get to 95% good answers!).

Ad Platform AI Optimisation For Offline Conversions

Moving our core ad channels (Google, Meta, Microsoft, Taboola) past lead volume optimisation to post-lead down the funnel events (discovery calls, appointments, sales) through improved Website > CRM > Ad Platform connections. Enabling lead quality optimisation, ROAS optimisation and AI-bidding for revenue/sales.

Daily & Weekly AI Reporting & Insights

Regular reporting from AI using CRM data – with insights and data shown decided by the AI each day/week to surface new/interesting insights.

Centralised Lead Processing & Data Standardisation

Updating of all ad channels, website & landing page forms and lead suppliers to go through one central lead processing flow which standardised data formats (ie. phone number), allowed better region allocation, and minimised rejection of leads from CRM + added error tracking & data backup.

Booking Calendar Management + Automated Reminders

Introduction of customer-facing booking calendars & integration with CRM plus automation of reminders to staff of upcoming bookings.

Plus tons of smaller tools, automation, workflows, and automated customer communication improvements!

What’s to come

I have a Miro dashboard with literally hundreds of ideas, tools to test, platforms to try and things that need to be solved. So plenty to work on.

Three projects I am excited to be working on next are:

  • Bad Lead Prediction – using historical data and then asking AI to predict a good or a bad lead so we can focus our call centre’s time on calling the best leads, and automate communication for those predicted to be lower quality/lower chances of converting.
  • AI Customer Support Tickets – helping our team manage a growing volume of customer support tickets with AI ticket replies, ticket escalation and hopefully automatically closing some tickets with no human needed (allowing them to focus on the harder/more time consuming tickets).
  • AI Customer Support Live Chat – implementing an AI live chat bot onto our customer support portal to help customers troubleshoot and solve problems automatically.

That’s it for now – if you’re working in performance marketing, AI or automation id love to hear from you so connect with me on LinkedIn.

Leave a Reply

Your email address will not be published. Required fields are marked *