# Using AI in fundraising – where it helps, where it doesn’t, & key challenges

- Link: https://fundraising.co.uk/2021/06/16/using-ai-in-fundraising-where-it-helps-where-it-doesnt-key-challenges/
- Published: 2021-06-16T13:46:40+01:00
- Author: Melanie May

**AI is helping fundraising in many ways, from improving targeting through machine
learning and algorithms, to engaging supporters and encouraging giving through voice
assistants and chatbots. But what do fundraisers need to know about the technology
and its subsets? We asked the experts.**

Firstly – it’s important to know what it is – and what it isn’t. The consensus is
that it’s still generally nothing like it’s portrayed in film and TV.

Rhodri Davies, Head of Policy at [Charities Aid Foundation](https://www.cafonline.org/),
says:

## Advertisement

> “The discussion is often framed as if we are talking about computers that have
> general intelligence at a level that matches (or surpasses) humans, but that is
> not really what AI means at all at the moment. Artificial General Intelligence(
> AGI) is the theoretical, sci-fi form of AI that we often think of, and most experts
> think we are nowhere near that (and it may be theoretically impossible). What 
> we do have now is computer programs with domain-specific abilities that match 
> or outstrip humans – sometimes in rigidly rule-based areas like chess or Go where
> computers have always been able to compete, but more recently in tasks that humans
> were traditionally good at but computers were bad at such as image recognition,
> and natural language conversation.”

#### AI in fundraising – how it’s helping

In practical terms for fundraisers, this means the likes of:

 * **Enabling donations** via conversational interfaces like Alexa and Siri, such
   as the [NSPCC’s Alexa skill](https://fundraising.co.uk/2021/06/09/5-ways-ai-is-being-used-in-fundraising/).
 * **Robotic Process Automation** (RPA): low-cost AI tools that can automate basic
   tasks (or parts of them) like diary management and data entry, giving fundraisers
   more time for their main jobs.

Matt Haworth, Co-founder and Director at tech social enterprise [Reason Digital](https://reasondigital.com/)
says:

> “One of the more realistic and practical application of AI for many charities,
> is actually less about the sexy exciting kind of above the line stuff where AI
> is interacting directly with fundraisers or with donors, and more about the mundane
> things it can do for you that otherwise keep you from using your precious human
> time stewarding donors and developing those relationships.”

 * **Generating content**. There are AI tools that can automate tasks such as producing
   written content for marketing, and creating natural sounding language.
 * **Analysis**. AI and Natural Language Processing are also being used in social
   listening tools and to analyse the performance of campaign and marketing collateral
   to help spot patterns, and make predictions.

Amy Wright, Director of Client Strategy at [Automated Creative](https://automatedcreative.net/),
which specialises in using AI to deliver performance and insight for social media
and display advertising campaigns, explains:

> “AI can tackle more than numbers. Unconventional datasets can reveal hidden correlations
> that help improve fundraising efforts. Images, copy, sound, or marketing data 
> can all be analysed – turning brand collateral into powerful insight for your 
> whole organisation.”

 * **Recommendation algorithms**. These use data to generate tailored recommendations
   to an individual – something we’re all used to in the consumer world, and are
   starting to be used by charities, for example on websites and with chatbots to
   promote the content most relevant for that person.

 * **Predictive analytics**. Machine learning can be applied to donor or prospect
   lists to spot patterns and make predictions to improve prospect research, or 
   gain a better understanding of existing supporter preferences and behaviour.

Graham Covington, founder and CEO of [Engaging Networks](https://www.engagingnetworks.net/)
and sister company [Accessible Intelligence](https://accessibleintelligence.io/),
which provides machine learning services to nonprofits, says:

> “The heart of machine learning is all about problem solving. In fact that’s what
> algorithms do – they’re set up in order to find a better a solution to a given
> problem. For example, trying to upgrade one time donors to monthly donors or increase
> retention of donors over a period of time.”

Using machine learning to get more from data and improve communications with prospects
and supporters is certainly one of the main areas where AI is proving of benefit
for fundraisers, and there are [many applications in CRM systems and the tech they link to](https://fundraising.co.uk/2021/01/13/ai-machine-learning-in-crm-systems-opportunities-benefits-for-charities/).

Chris Paver, COO at tech firm [Dataro](https://dataro.io/), which specialises in
using machine learning to help nonprofits with their fundraising says:

> “Machine learning is the next frontier for fundraisers when it comes to making
> the most of the data that exists in their CRM and making smarter fundraising decisions.
> This is an area where nonprofits have a real opportunity to achieve significant
> improvements on the fundraising outcomes they’re currently achieving and raise
> more money across all of their fundraising programmes.
>  
> “The techniques most charities currently employ, like RFM analysis, only take 
> into account a very limited snapshot of information about the donor and their 
> relationship with the charity. Machine learning however allows charities to take
> a holistic view of all the data they’ve been diligently collecting in their CRM
> and actually put that to work, so they can start to use all the information they
> have to its full potential and figure out who they should be talking to about 
> what.”

More can be read on how AI is being used in fundraising [here](https://fundraising.co.uk/2021/06/09/5-ways-ai-is-being-used-in-fundraising/).

#### What to watch out for

But of course, as with any technologies there are some potential challenges and 
issues for fundraisers to be aware of. For one, AI and machine learning are only
as good as the data they’re working with and machine learning algorithms require
large volumes of data in order to learn patterns and behaviours, and create models
based on this. Having enough data and having it in a usable, useful state could 
be an issue for some charities.

Steve MacLaughlin, Vice President for Product Management and Senior Advisor to the
[Blackbaud Institute](https://institute.blackbaud.com/) says:

> “If we go back to that definition of AI – that it’s all about learning from data–
> then that can present some challenges. AI is only as good as the data it learns
> from. Bad AI and Bad Robots in the movies likely all came from bad data along 
> the way. This is why the quality and appropriateness of the data used is so important.”

The importance of integrating all data sources into a single supporter view for 
managing relationships and compliance has been much discussed over the years, but
it’s just as important for the performance of machine learning applications.

Covington comments:

> “As charities start racing more and more towards using machine learning applications,
> they’re going to start realising the importance of having integrated data around
> an individual, and being able to make sure it’s good, appropriate, relevant data
> and it’s permission based. And that’s going to be an interesting challenge for
> a lot of organisations, because many don’t necessarily have their data organised
> in a useful way.”

And, clever though AI and its subsets are, in many ways they still can’t replace
human intelligence and all activity needs to be overseen, so it’s also important
to keep people involved in the process.  While visual recognition for example can
be used to interpret images, although it will correctly interpret what’s obvious–
such as water, trees, and sky, it will be less good at understanding context and
working out that people standing in water might be part of a volunteering activity,
or affected by a flood, meaning applying human intelligence to this kind of task
is still critical.

Human intelligence and involvement is also important when using AI and machine learning
to interpret data and create models and predictions to use in campaign, one reason
being because data records need to be correctly tagged for machine learning to deliver
coherent outcomes when looking for patterns.

Paver says:

> “A common pitfall to watch out for is the idea of a set and forget model – where
> you conduct a modelling exercise and land on a model that you think is going to
> be predictive for a particular outcome, such as giving to an appeal. You can’t
> just do that process once and assume that model is going to work forever. There
> are going to be changes in the data and new information will become available 
> that might mean that model becomes less and less effective over time. The exact
> same thing applies to the predictions themselves, so it’s important that you’re
> constantly updating these too, because they will also go stale as new information
> comes to light, and that donor continues to interact with the charity.”

Wright agrees, saying:

> “There are things machines are brilliant at – automating repetitive processes;
> speeding up manual tasks; and allowing you to work at much greater scale. But 
> they are literal, and work best when deployed strategically. At Automated Creative
> we use machine learning to augment human creativity in the marketing and fundraising
> process – not replace us completely!”

Aside from these practical considerations there can also be moral and ethical challenges
to deal with, particularly where machine learning algorithms and the risk of biases
are concerned, as Nigel Magson, CEO of data and insight agency [Adroit Data & Insight](https://adroitinsight.com/)
explains:

> “One of the ways you build algorithms is to train them on data sets to recognise
> things – whether it’s visual recognition, or a data combination recognition  –
> but because they learn by being trained on big data sets, they may include any
> biases, such as age, gender, and race, that were in those data sets, so the ethical
> point here is to understand how your model might be discriminating and to ensure
> therefore that any use of algorithms does not result in bias against particular
> groups of communities.”

While this is a small look at a big topic, what’s clear is that while AI is still
really in its infancy in fundraising, and there are some hurdles and limitations
for fundraisers using it to overcome, its applications – and potential benefits –
are practically endless. From predictive analysis and modelling through to generating(
and analysing) content, enabling giving, and providing supporters with relevant 
information, it’s certainly got much to offer.