SpeechSage

03/10/2026

09/24/2025

SpeechSage is at our first trade show!

We are pleased to announce that we will be at the South Dakota Association of Healthcare Organizations’ annual convention!

Stop by booth #21 tomorrow, September 25, during the Exhibitor social from 4-6 PM for your chance to win our door prize game, snag some awesome merch, and satisfy your sweet tooth with candy. We'd love to meet you!

Unrelated pic as we have not been to a trade show before

06/30/2025

When potential customers call your business, timing matters more than you might think. Recent studies reveal some eye-opening patterns about customer phone behavior that could impact any business.

Over 80% of callers who reach voicemail simply hang up rather than leave a message. Meanwhile, most people expect to wait around 5 minutes when calling a business, but their patience often runs shorter than that expectation. Small businesses report missing over 60% of their incoming calls, and here's what's particularly telling: two-thirds of customers who don't connect on their first attempt won't try calling back.

Consider your own habits when you need to reach a business quickly. If you can't get through, you probably move on to the next option, right? Your customers likely do the same thing.

This isn't just about convenience – it's about how modern consumers make decisions. In today's fast-paced environment, accessibility often determines which businesses get the opportunity to serve customers and which ones miss out entirely.

The phone remains a critical touchpoint for most businesses, serving as both a customer service channel and often the first real interaction someone has with your company. Understanding these calling patterns can help businesses make more informed decisions about how they handle incoming communications.

At SpeechSage, we've seen how consistent phone availability can transform customer relationships. If you'd like to explore how professional phone answering might fit into your business strategy, we'd be happy to discuss the possibilities.

06/18/2025

Why would you want to integrate Voice AI into your healthcare business?

See the use cases in the short blog below!

The AI industry churns with hype, and I mean a lot of hype. Some observers compare today’s AI fervor to the early-2000s dot-com bubble, and…

05/27/2025

Ever heard of federated learning? It’s a machine learning concept that could reshape how we handle sensitive data, especially in healthcare.

➡️ Question 🙋: What is federated learning? Answer ✅: Federated learning is a technique that allows "local" models to collaborate, to train an ML model on a private dataset without ever sharing the underlying private data. What does this mean?

🧠 Intuition: Let's say Bob, Alice, and Steve all have their blood pressure data on their smartwatches. They want to be able to generate a prediction model that will tell them when they are at risk for a stroke. However, Bob does not want Steve to know his heart rate information (Steve is a UND fan), and Alice does not want either of them to know her blood pressure information. But if they use federated learning, they can train a small ML model on their smartwatch, and then upload the ML model to an agreed-upon website (server), without uploading the actual blood pressure data. This is the concept of federated learning.

➡️ Question 🙋: Why is this necessary? Answer ✅: A few reasons:

(1) Current encryption techniques are not guaranteed to be secure; the best data security is locally, on your own device, not on the Internet.

(2) This allows for better ML model training. Oftentimes, when training on healthcare or sensitive data, the sensitive parts are redacted. This is less data to train on, which one can presume lessens the efficacy of the model.

➡️ Question 🙋: What is the future of this technology? Answer ✅:

Currently, there is still ongoing research and development for the use of these models. However, as artificial intelligence machines learned large language models (AI ML LLM, whoa that's a lot) get more performant, at smaller sizes, and as devices get more powerful in general, the efficacy of this method will become more and more feasible.

➡️ Question 🙋: What domains will this be good for? Answer ✅: Any domain that uses sensitive information, including but not limited to, healthcare, finance, and internal organization AI-capable tasks.

🧠 Intuition: A little intuition for clarity's sake; why this is so attractive is that it will likely allow for much larger datasets to train models on. Instead of needing to "give your data" to a possibly unknown source (which can have many downsides), you can be assured that your data never leaves your device, but while still getting the benefit of the data of everyone else's data in the resulting model. A win-win for all.

Follow SpeechSage to see how we are trying to help Healthcare businesses utilize AI in their day-to-day operations!

Here are some resources for further information about federated learning:

https://en.wikipedia.org/wiki/Federated_learning

https://www.intel.com/content/www/us/en/research/news/federated-learning-protecting-data-at-the-source.html

https://www.geeksforgeeks.org/collaborative-learning-federated-learning/

Your All-in-One Learning Portal: GeeksforGeeks is a comprehensive educational platform that empowers learners across domains-spanning computer science and programming, school education, upskilling, commerce, software tools, competitive exams, and more.

05/17/2025

Sometimes it can seem like there is too much hype in AI.

We hear "AI this" and "AI that" and "AI will change everything".

Beneath the hype there is some truth. But what is the actual truth? We have been promised 'AGI' for a few years now, while our state of the art AI makers have problems making sure they don't hallucinate (https://openai.com/index/sycophancy-in-gpt-4o/, https://apnews.com/article/elon-musk-grok-ai-south-africa-54361d9a993c6d1a3b17c0f8f2a1783c)

AI will be very beneficial. It already is proving so in particular domains. But like any new technology or relationship, trust must be built up. The blog below talks a little bit more about that.

https://medium.com/.vandal/ai-has-too-much-hype-430a1a4ae8f3

05/16/2025

Interested in considerations for utilizing AI in your small business?

This is a concise write-up about some thoughts before taking that step.

https://www.forbes.com/councils/forbesagencycouncil/2024/08/01/a-guide-to-ai-voice-agents-for-business-owners-and-leaders/

05/15/2025

Some businesses may have apprehension about using AI in their practice - that is understandable!

New technologies take time to get used to. What helps with getting used to a new technology is understanding of how it works. The unknown is what causes fear, not the known.

A company (not sponsored) that has created a voice AI technology that sounds very realistic is Sesame. They also published an open-source version of their model and a paper describing how it works.

Try out the demo here!

At Sesame, our goal is to achieve “voice presence”—the magical quality that makes spoken interactions feel real, understood, and valued.

05/14/2025

Have you ever wondered how a large language model works?

This article is a great write-up of how they work on the inside.

⬇️ Article below:

From zero to ChatGPT

05/13/2025

Language is information being conveyed.

When speaking, different languages convey information at different rates!

For example, the Japanese language has about 640 syllables; these syllables convey information at the rate of about 5 bits per second; contrast this with English (that has over 6900 syllables), and this languages conveys information at the rate of about 7 bits per syllable.

One study found that the average rate of information transmission across all languages was about 39 bits per second - this would be equivalent to approximately 30 words per minute typing.

Read more here!

No matter how quickly you speak, you still share the same amount of information