Using AI with SMS
“AI and SMS?!” I hear you say. How can cutting edge technology impact something as old and crusty as SMS? Well, in plenty of ways it turns out.
If we take a step back and look at the key attributes of both technologies, it’s pretty obvious that they can be used together – AI allows us to achieve things in a more productive way and SMS is still the most effective, universally accessible communications technology, with 90% of messages being read within 3 minutes of being delivered, according to Forbes. Taken together, SMS provides an efficient interface for accessing powerful algorithms and models that can help people to communicate with machines.
AI is an unavoidable term these days (and yes, we are adding to it with this article), but it’s used very loosely and covers a lot of capabilities. To understand how we might be able to use it we need to understand what it is and what it can do for our businesses. Marc Andreeson provides a good, short description of what AI is:
“The application of mathematics and software code to teach computers how to understand, synthesise, and generate knowledge in ways similar to how people do it.”
An even shorter way to say it is that AI takes lots of input and generates a usable output for us. It can do this in several ways:
Descriptive AI – A capability which allows items to be categorised by a machine, without human input, like telling the difference between pictures of cats and dogs.
Predictive AI – Using statistical models, predictive AI’s allow a software program to predict what might happen in the future. These are used to develop things like movie or product suggestions, based on our previous activity.
Prescriptive AI – Using optimisation and simulation algorithms to consider and range of outcomes, prescriptive AI’s make suggestions about what we should do. These allow companies to assess a variety of future outcomes and the factors that influence them. It’s a good way to evaluate various scenarios, understand risks and potential impacts.
Generative AI – This is what all the recent fuss is about. Generative AI uses things like Large Language Models (to generate text, like ChatGPT) and Diffusion Models (to generate images, like Midjourney) when given a prompt of some kind. These help us to get detailed answers to questions very quickly and (of course) to generate silly images of our colleagues:
Whilst we’re probably not going to spend our working hours frankly disconcerting pictures, we should be considering how AI can help both our businesses and our customers, and how best to provide people with access to the outputs of these AI models.
How can AI + SMS help my business?
Given the variety of outcomes that AI can support, there are probably things that you’re already doing today that could benefit from artificial intelligence and SMS, as well as some new things that you can do which either weren’t possible or cost-effective before now. Here are a few ideas:
Automating responses – AI-powered chatbots can use SMS to provide accessible and instant responses to common queries. These can understand natural language to engage in conversations which provide helpful information, or gather basic details from people to provide to someone who can take further action with the helpful context provided up-front.
Personalised messaging – AI algorithms can analyse large reams of data to expose user preferences in order to generate personalised SMS messages. These allow you to tailor your recommendations or notifications, based on the interests and behaviours of your customers. This isn’t just for product organisations – The SMS Works is supporting healthcare providers who are using AI to determine the best way to help patients, many of whom have significant technology challenges. They are using data on previous behaviours and personal preferences that allow healthcare providers to smartly schedule appointments and help patients to avoid missing them by sending smart reminders.
Sentiment analysis – AI can be used to analyse the sentiment of incoming text messages to determine whether messages express positive, negative or neutral sentiment. This could be used to trigger onward customer support activity, where someone needs to intervene to help, or allow you to monitor brand sentiment, as part of an NPS survey, for example.
Intelligent routing – AI can optimise the routing of messages, based on factors such as time zone, to dynamically determine the best route for a message to take (e.g. SMS or email) or the best timing for delivery, to ensure that messages are delivered sensitively and support customer preferences or optimise conversion rates.
Language translation – AI-powered language translation can be very useful for assisting communications between people who speak different languages. These can be especially useful in customer service situations, to provide sensitive help where a language barrier can add stress to an already difficult situation.
Spam filtering – AI can help SMS providers like us to filter out spam messages in SMS platforms. By analysing patterns, keywords and sender reputation we can identify and block unsolicited or phishing messages, improving the overall experience and security for the recipients of messages sent from our platform. You might use these sorts of filtering algorithms to highlight undesirable activity on your own platforms.
Optimising content – the limit of 160 characters for each SMS message means that, if your message is slightly longer that 160 characters, you’re going spend twice as much to send a few more words; content that may well be redundant. Large Language Models like ChatGPT can be used to modify your messages, to reduce their length without losing the information or tone of your message. The SMS Works has a native AI Message Optimiser, that can automatically target messages that are just that little bit too long, and shorten them for you.
How can I add AI technologies to my existing systems?
Most companies employ specialist software that meets some or all of their specific needs, as defined by how they run their business and serve their customers today. When it comes to AI, a lot of people might consider these systems to be ‘legacy’, and not able to cope with running things like natural language processing. So, how do we integrate these capabilities?
‘Integrate’ is the key word here. Much as you may already have integrated with an API like ours to send SMS, companies are beginning to offer AI-as-a-Service platforms that allow us to layer these capabilities on top of the systems we’re already using.
For example, the version of ChatGPT that we access on our web browser or phone uses billions of publicly available web pages and scholarly articles to generate the answers to the questions that we pose, but it doesn’t have access to our internal process or system documentation. However, OpenAI provides an API that we can use to add ChatGPT’s capabilities to our documentation, which we can use to build a natural language Q&A chatbot that can consume questions and provide answers via SMS.
All of this can be done today using the same technologies that you’ve used to integrate SMS already, whether that’s something as simple as cURL, PHP or even Python. You don’t need to know how the AI works, just the right way to deploy the API it to assist your staff or customers and how to integrate it into your systems.
Which companies offer AI offerings that I can integrate today?
We’re really at the beginning of a Cambrian Explosion of AI software providers. This includes the large tech companies (listed below), with the resources to access large data sets and fund big language models, the plethora of startups that use these capabilities to fill every niche imaginable, as well as open source software that is available without charge and supported by the community. Your engineers might want to start with seeing what the bigger companies offer, as their models are more often than not what startups are using under the hood.
| Company Name | Website | Description |
| OpenAI | openai.com | OpenAI offers a range of AI models and tools, including the GPT-3 language model, which can be accessed through their API. It enables developers to integrate natural language processing, text generation, and other AI capabilities into their applications. |
| Google Cloud AI | cloud.google.com/ai | Google Cloud AI provides a comprehensive suite of AI services and models, such as Vision AI, Natural Language API, and AutoML, all accessible through their API. Developers can leverage these capabilities to enhance image recognition, language processing, and more. |
| Microsoft Azure Cognitive Services | azure.microsoft.com/services/cognitive-services | Microsoft Azure Cognitive Services offers a wide range of pre-built AI models and APIs that enable developers to incorporate vision, speech, language, and decision-making capabilities into their applications, simplifying the development process. |
| IBM Watson | ibm.com/watson | IBM Watson provides a set of AI-powered services, such as language understanding, visual recognition, and speech-to-text, which can be accessed via their API. These services allow developers to integrate cognitive capabilities into their applications with ease. |
| Amazon AI | aws.amazon.com/ai | Amazon AI services, available through AWS, offer a wide range of AI capabilities, including computer vision, language understanding, and speech recognition. Developers can access these services through APIs to enhance their applications with AI functionalities. |
Conclusion
The use cases above demonstrate the versatility of AI in SMS applications, providing your customers with more functional, efficient and secure interactions in various scenarios, and in a more scalable way than you may be able to manage today. We’ve come up with a few ideas of how we might use AI at the SMS Works just in the process of researching this article. We hope this helps you to do the same.
