Intro to AI for curious not-for-profits.
Video Transcript.
“Artificial Intelligence is important because it has the potential to significantly improve processes, create new ways of working, identify efficiencies, and complete tasks in a way far beyond what human beings can currently achieve.
In a business context AI could help us work smarter and faster on the activities that really matter to our success. It could also free our staff from some of the more repetitive and low value activities to focus on activities of greater value.
Different types of artificial intelligence.
The 2 most common types of AI currently are Machine Learning and Deep Learning.
Here’s my explanation the difference types of AI:
Let’s think about window blinds. A simple set up would tell the machine that the sun is shining between 6am and 6pm and that the blinds should come down during this time.
Machine learning would add a light sensor to the machine, program the sensor to identify when light was brighter and then depending on the time of day lower the blinds when the light was within certain ranges of brightness.
Deep learning would add additional sensors and program the computer with a range of scenarios that indicate my preferences for when the blinds were lowered. The machine would continuously monitor the light, learning and adjusting for different combinations of sun, clouds, rain etc and adjust the blinds up and down based on its reading of the level light, my preferences, the weather forecast and the time of day.
Currently most AI tasks fall into the following groups:
Visual tasks involve processing and interpreting image-based "things" eg: the face recognition on your phone
Linguistic tasks involve handling and using language-related functions eg: Amazon Alexa , Apple Siri or Google Assistant
Prediction tasks involve sophisticated processing and interpretation of information and data to make recommendations ag: Netflix suggesting movies you may like
Examples of AI in action today.
We currently have access to a range of AI solutions in our businesses. Some examples are:
Medical researchers using AI to prevent vision loss and scan for cardiovascular disease.
Researchers using AI to improve communication with non-verbal people with disability.
An AI doll for use by older people to provide companionship and reduce loneliness.
Mental health researchers using AI to predict when someone may commit suicide.
And, child protection services using AI to predict children at risk.
The point to note is that the technology to complete these tasks exists with us today and that open-source platforms like Amazon Web Services allow us to pro-actively design and implement AI solutions now.
Artificial intelligence in fundraising.
Lets look at one real world use case from the fundraising space in more detail:
An online article in Forbes magazine from Nov 2020 talks about some of the successes that AI support.
A Director of Development at Arkansas State University speaks of his success with AI solutions that helps manage a portfolio of donors by identifying the right time to make contact, drafting emails that will engage and keep the relationship moving, provide weekly insights into the portfolio and even plan efficient travel routes and arrange meetings when they are on the road.
The success of these AI solutions has been:
Allow the fundraiser to actively manage a donor portfolio that is 66% larger than before
Support a 37% increase in high impact actions in the first six months of use;
And a 175% increase in funded proposals.
This is a great example of the potential AI has in supporting fundraising teams to deliver results.
The challenge that many fundraising organisations face is that the range of AI solutions available rely heavily on large volume, quality data, entered in a timely manner. Any predicative AI outputs are only as good as the inputs that it does the analysis on. For most time poor fundraising staff the entry of data at the end of a long day of donor meetings is a secondary concern. We need to change this.
An AI solution for modern for-purpose fundraising.
The following solution is proposed as a way to use AI to address this challenge by:
Improving the quality and timeliness of data captured and inputted into the CRM system post donor meetings.
And building additional donor relationship management capability in the fundraising team.
The proposed AI solution is Betty the donor meeting coach. Betty will be programmed to:
Think
Speak
Analyse
And Coach.
At a simple process level Betty will:
See a donor meeting has been booked
Schedule a phone call
Call the fundraiser
Complete a structured Q&A conversation
Prepare a draft report
Accept fundraiser corrections
Add all meeting details and notes into CRM
Schedule donor follow up
Assess fundraisers meeting skills
Lets look at a short example of one task Betty would complete:
When Betty calls a fundraiser she will ask a range of questions for the fundraiser to answer. One of those could be “Did you ask the donor about their next gift?” to which the fundraiser will provide a response.
Betty will be programmed to analyse the response for key pieces of data that indicate the completeness of the response.
Betty will then label the data and then if all the required data points have been supplied assess the response as complete and move on to next question.
If the response provided was incomplete and missing required data, in this case there is no project or product provided, Betty would ask a follow up question and allow the fundraiser to provide a response.
In this way Betty will ensure that all relevant data is collected and inputted into the CRM. Betty will also note the overall level of completeness of the fundraisers responses and if training is required send an email to the fundraiser with some relevant learning and development content.
The outcomes would be significantly higher quality of data collection and accurate labelling in the CRM system, and an increased level of capability in the fundraising team.
Ethics in AI.
Any proposed AI solution must address ethical concerns around AI. When we design a solution like this, we use a lot of data, in some cases private and personal data collected over time, and program a machine to use this data to provide insights, identify data points as “good” or “bad”, train the machine as to what is “right and “wrong”.
As solid and robust as our programming might be a machine can never learn human level ethics and morals, it will be ways guided by the person who programs it and the data it is given to work with.
Bias in artificial intelligence.
Bias is an example of the challenges that AI faces. Based on the data used in its programming some AI solutions have shown bias toward males over females and European faces over African and Asian faces.
In 2019 an AI solution designed to assess and approve credit applications was shown to be bias towards women and gave men faster responses and higher credit limits, even when they had the same financial details. This issue occurred because the historical data used to program the AI had the same bias, and hence the machine learnt to make decisions in the same way.
For the Betty solution there are some challenges to be aware of including:
Bias - As the Betty solution will assess a fundraisers performance by listening to their responses over a phone conversation, we need to make sure that answers from people with a diverse range of accents, males and females, loud voices/soft voices are all able to be heard and that the solution hears the data they provide in the same way.
Donor privacy and cyber security – making sure donor data is protected and safe as it flows across multiple cloud-based products
Component integration – reading and interpreting data, making phone calls, text to voice to text, chatbot functionality, natural language processing etc the solution relies on a range of components that need to be successful integrated.
The things we don’t know yet - we will uncover unknown challenges as we create new ways of working.
I hope this short overview of AI in the not-for-profit sector was useful and has help you to start to think about how AI might be used in your business.
If you would like to keep the conversation going please message me of visit me at insight & foresight
Have a great day.”
Script prepared as part of RMIT “Developing AI Strategy” Course. See credential here https://www.credly.com/badges/c647c2ea-e443-4af2-bf9a-a01940fe8211
