Interview with Said Kassim, co-founder of Echo 

Taaher Robbani, Solicitor in Birketts’ commercial and technology team, recently sat down with Said Kassim (Said), co-founder of Echo, a platform built to support customer service teams during and after calls. Echo listens to live conversations, helps agents find the right answers quickly, and takes care of the follow-up work by syncing reports straight into the CRM. Said shared how the idea came together, how the tech works, and why Echo is designed to fit into any team’s workflow without needing complex integrations. 

Can you explain what Echo does and how it supports customer service teams?  

When we started Echo, we actually began with something pretty simple: agent assist. The idea was that the app would listen in on customer service calls and help the agent in real time, basically like having a sidekick who is always there, ready with the right answer. No more digging through folders, asking line managers, or putting customers on hold. It is all backed by an intelligent knowledge base that understands the conversation and gives the agent what they need, when they need it. 

Originally, we wanted to build a full voice AI that could replace the agent altogether. But when we started speaking to potential clients, about a year and a half ago, there was a lot of pushback. People still wanted that human voice, that authentic tone that represents their brand. Plus, there were concerns about AI hallucinating or giving the wrong answers, which led to us pivoting. Instead of replacing the agent, we focused on supporting them. That is how Echo became a real-time assistant. It listens, understands, and helps the agent complete the call faster and more accurately, keeping the customer happy and the agent confident. 

But then, as we kept talking to customers, another need came up. After-call work. Agents were spending anywhere from two to ten minutes writing up reports after each call. And in some cases, they had to send those reports to a third-party client. For example, company X might outsource their customer service to company Y, and our software would sit with company Y. But company X still wants those reports instantly so they can take action. 

So, we built another layer on top. Since we already had the transcript and context of the call, we automated the reporting. Now, as soon as the call ends, Echo syncs everything with the CRM.  

The agent does not have to do anything; they just focus on the conversation. By the time the call ends, the report is done. Everything is submitted. Job done. 

It is a huge win in terms of efficiency, cost savings, and just making life easier for everyone involved. 

Where did the idea come from?  

The idea for Echo was not originally mine. At the time, I was working on a different product, a knowledge base for enterprise teams. The problem I kept running into was that every time I joined a new company, I would waste hours trying to find information, book time with senior staff, and get up to speed. I thought, “What if I had a sidekick in Slack or Teams that could just answer my questions instantly?”. 

That is when I started building. And that tech ended up forming the foundation of Echo. Then I met my co-founder, Idris, at Hulm Club, a co-working space in Shoreditch. We got talking, and he told me about his experience working as a customer service assistant on the NHS 111 line. He faced all the same challenges we were trying to solve: pauses during calls, scrambling for answers, and the pressure of getting it right. He said, “Wouldn’t it be cool to have a tool like Echo?” And that is how it all came together. 

You can really see the value of something like Echo in that it can provide you with quick answers, which could be crucial in decision-making for a call handler. 

What AI models underpin Echo’s service?   

Echo started as a knowledge base tool, built using a RAG (Retrieval-Augmented Generation) system. The idea was to pull in data from all kinds of sources, for example, text, audio, images, structured or unstructured and centralise it. We built connectors for everything from Microsoft Teams to databases, so no matter where the info lived, it could be indexed and made searchable. 

From there, we layered on a real-time agent assist product. The challenge was accessing live audio from softphones, which do not usually expose that data. So, we built a Chrome extension that taps into the browser’s audio layer, letting us capture the conversation in real time. That audio is then transcribed using speech-to-text models like Whisper. 

Once we have the transcription, Echo can do a lot: it pulls answers from the knowledge base, shows them to the agent mid-call, and even analyses sentiment in real time. If a customer’s getting frustrated, Echo flags it. It also helps with Q&A, training, and reporting, basically turning every call into a data-rich resource. 

The key is that everything hinges on real-time transcription. Once we have that, we can support agents live, help businesses improve, and adapt to any industry without needing deep integrations. 

How does Echo protect the data that it processes?  

While we have not worked directly in the medical space yet, we have always built Echo with GDPR and data protection in mind. So, before any data goes to a language model, we run a redaction step using named entity recognition. That helps us identify and mask personal info like names, phone numbers, and other sensitive details. 

We use NLP models to do this, some off-the-shelf, some custom, so we can reliably spot what counts as personal data. On top of that, we are cloud-native, but we have also made Echo deployable locally for clients who want everything to stay in-house. Most are happy with cloud, though, since we use trusted providers like Azure, AWS, and Google Cloud, all of which have strong security and encryption standards. 

As for the language models themselves, we are careful there too. We do not send data to OpenAI directly; we use OpenAI through Azure, which has specific terms that prevent your data from being used for training. That is a big deal, especially in AI contracts where data privacy is a sticking point. We also use different models for different parts of the product. For example, when we are generating reports to submit to CRMs after a call, we might use OpenAI via Azure. But for real-time tasks like pulling out phone numbers or emails, we have found Anthropic Claude to be more accurate. So, we mix and match depending on what is needed: speed, accuracy, or cost. 

The idea is to eventually fine-tune models for each domain we work in, but even now, we are already tailoring the tech to fit the job. It is all about using the right tool for the right part of the pipeline. 

What would you say is Echo’s USP?  

What makes Echo different is that we did not build it for one specific industry. A lot of other tools go deep into one niche, like the NHS or care homes. We went broad. Echo works across any CRM, any dialler, any knowledge base, even if it is just a Google Doc or a stack of paper. That flexibility means that it can be used in sales, real estate, dental practices, you name it. The tech stays the same, and it just works. 

We did not want to build custom integrations for every setup. So, we made Echo integration-free. As long as there is a form or a field, Echo can read it and fill it. That has been our edge, building something that is simple, adaptable, and ready to plug into whatever system a team already uses. 

The Birketts view 

Echo’s approach to real-time support, automation, and data protection reflects the kind of innovation we are seeing across many sectors, including those we advise at Birketts. As businesses adopt AI tools like Echo, legal considerations around data privacy, contract terms, and integration risks become increasingly important. 

At Birketts, we help clients navigate these challenges, whether it is drafting robust SaaS agreements, ensuring GDPR compliance, or advising on the use of AI in regulated environments. Echo’s flexibility and integration-free design are particularly relevant for clients looking to scale without heavy infrastructure changes. 

As AI continues to reshape customer service and operational efficiency, Birketts is here to support businesses in making informed, legally sound decisions.