As generative Artificial Intelligence is rapidly increasing its ability to boost content-creation, there is an associated concern about the potential impact of automation on the availability of high-quality online data. Spoke AI, a Berlin-based startup, is preparing to utilize generative AI within the parameters of businesses with the goal of assisting their workers in dealing with the influx of communications they receive through a variety of external platforms by doing automated summaries.
The ambition of the startup ultimately is to provide AI productivity applications that can be utilized by personnel in the entire company. To begin with though, it is concentrating particularly on AI aggregation and summarization software for project administrators.
The idea is that this group of desk workers tends to use many different third-party software programs, such as Slack, Jira, Github, Miro, Figma and Notion, which requires additional help to manage the large number of disjointed notifications and messages coming in. Later, once the company has refined its technology and compiled new sets of training data, the goal is to gradually roll out products designed to give aid to all sorts of information workers.
This week, Spoke AI offered access to its first beta tools and according to them, they have around 20 trial-users with a waiting list of over 500 organizations that want to give their technology a go. Several businesses, from small ones with 10 employees to ones with a whopping 250 workers, have expressed their interest. Furthermore, Spoke AI is creating a feature for larger firms.
The Q1 2021-founded startup has generated significant attention from investors and is now revealing a â¬2 million (~$2.1M) pre-seed financing round led by the Northern European-based byFounders fund. Possible Ventures also contributed to the round. On top of that, the team was awarded a grant from the European Regional Development Fund through Berlin’s regional development bank, IBB. In addition, the group had already earned some angel funding to develop their MVP before the pre-seed.
Max Brenssell, one of the co-founders, states that the utilization of AI intends to decrease the amount of distractions in their day-to-day work with many instruments and websites they use. Firstly, this is designed for product managers whose task involves handling multiple tools. AI serves to aggregate, prioritize, and condense the exchange of information between the different tools.
The new business’s introductory package – marketing it as a “workplace operating system” – contains a search capability that can draw data from numerous external applications, for instance chats and tickets that the user has been named in, to establish a “clever inbox” experience with AI-constructed “contextual summaries” on top. To start off, they are planning to run trial runs of this with some of the greater corporations.
Swell is putting forth a search feature (called a Generative Knowledge Base, or “astute search”) as a program expansion, giving clients the capacity to go across pertinent tools in their so-called “savvy inbox” so as to get answers in an abridged structure as opposed to linking out to obsolete webpages.
Spoke AI is supplying their automated summarization as a Slack add-on to early users – the intention being to give a service where the intended customers spend a lot of their time, as opposed to requiring them to post what can be confidential business information to an unfamiliar website.
The use of AI to pull together relevant yet scattered information from different sources into one centralized place is not a new concept. What provides an extra value something is that is the capability for the AI to generate contextual summaries which helps to maintain the original context that can be easily lost when the data is taken away from its original environment.
Spoke AI emphasizes security and privacy as one way to distinguish itself from traditional methods of augmenting efficiency by collecting information.
Brenssell states that the real key to success for this particular instance lies in the summarization that works effectively and in a secure manner. To guarantee this, they use pre-trained language models, such as those of OpenAI’s GPT. However, we do a lot of work ahead of time as well as afterwards to anonymize the data and make sure it is properly safeguarded from a security and privacy standpoint.
He states that Spoke AI sees this element potentially increasing another revenue source – which could be if they offer data concealment as a service (through an API) to other organizations that need to apply AI models such as GPT to their own unique data sets, alongside offering AI condensation.
Brenssell has also indicated the possibility of transforming the principal summarization abilities into an application programming interface (API) so as to make money off of the system.
The company is currently providing its summarization technology in the form of a free plugin for Slack. Initially, this will be available as a subscription-based service with different rates dependent upon the amount of integrations, security measures, etc.
Precision is clearly a fundamental part of the company’s promise. If the summaries it makes are not correctly grasping the idea behind the notifications sent out, Spoke AI could end up being a massive burden on the people it is intended to help, instead of reducing their workload.
Brenssell mentions that the beta product includes feedback loops so that users may review the automated summaries and aid in the development of the product. They are also aiming to be transparent about what inputs the AI utilizes to make a summary, to give users the opportunity to understand the process. Earning trust through transparency is a main objective.
He points out that people have a common query, which is how they can believe what they are seeing. What they have done is to provide a transparent source of information that backs data in the summary, and give the user an option to explore the details further if they wish to.
Contemplation may be raised concerning whether AI summarization is sufficient to constitute a profitable enterprise on its own. Alternatively, maybe existing powerhouses of productivity, like Microsoft, which has put resources into OpenAI’s ChatGPT conversation passage as a generative AI, and has spoken about adding the technology to a collection of its own applications, can contribute further handy features. Google already employs AI for specific ‘intelligent’ functionalities, such as in its email service, however these have been of variable outcomes.
Brenssell expects the main contenders like Notion, Slack, etc. to introduce AI in their products, which is positive. They are tackling an issue that affects product managers especially, since these professionals have to use a large number of different tools daily. We can observe a rise in the amount of specialized instruments used in other areas as well.
Eventually, all of these tools will have the ability to exchange data and information, and this is what we found to be of big value, as far as developing a summary layer that can accommodate the different tools in a manner that makes users more productive while still maintaining the general idea of what is happening on the system.
Many corporations that collaborate with Jira from Atlassian tend to opt for Notion as opposed to Confluence due to its greater ability to adapt to their documentation. Furthermore, tactics such as these are becoming increasingly common as there is now an increased variety of efficient tools available for individuals to pick and choose from that best fit in with their team, process and business. Hence, that’s the current state of the market. We are depending on that accordingly.
Does it make sense to use generative AI to manage information overload when automation is likely to make it much harder to filter the important information from the useless data? It’s almost like creating an AI arms race that never ends, which does not often have a happy ending according to films.
Brenssell states it’s an “interesting point” and expresses curiosity on where generative AI tools will go in the upcoming years. He believes the risk of using AI in the business communication sector is low, while using it for tasks such as copywriting and outreach emails is more desirable.
He went on to say they have noticed a lack of use of generative AI when it comes to examining how teams in different companies can perform more effectively.
He recognizes that some services already provide summaries concerning features within their products, like Notion and Intercom. Additionally, there are mailbox aggregators which have the capability to bring up WhatsApps, iMessage and even business emails into a sole inbox. Nonetheless, he claims that the summary element is what may turn out to be a particularly differentiating factor.
He believes that the team’s commitment to data privacy, ethical data use, and creating a product with reliable summaries will help them stay one step ahead of any competitors who might use the same artificial intelligence model. He also noted that creating summaries is not as easy as developing content using AI. Clearly, the technology behind it is advancing. We’re staying up to date with these changes. In order for us to have an advantage on the AI front, we believe that by constructing the right components, such as data anonymization, is where we will be most successful.
Then the focus is on creating a seamless user experience when working with different products. We don’t want to force users to switch between products, like from Slack to a new tool, as that creates a lot of difficulty. Thus, we are striving to make our summarization fit into existing workflows and have a cohesive approach.
The company is utilizing a mix of AI models and tailored training for the summarization with Brenssell noting that it is “supported by a mix of adjusted pre-trained language models (for example, Luminous, GPT-3.5), in-house hosted open-source technology (e.g. GPT-J, BLOOM, technologies created by Microsoft and Stanford), and custom models trained in-house (e.g. for Named Entity Recognition, PII Detection, Data Pseudonymization, Question/Answer Identification, Semantic Search)”.
He is telling TechCrunch that although Natural Language Processing technology is becoming commoditized, they still feel it is important to differentiate. They think this can be accomplished by emphasizing data privacy, responsible Artificial Intelligence and augmenting human nature instead of automating it. Establishing faithfulness with users is essential and data processing both before and after that improves both security and the user experience is necessary. We are anxious to rapidly examine particular purposes with available technology and then commit to creating private data collections based on implicit and explicit reaction, modifying and teaching our models on a deeper level.
Casper Bjarnason, investor at byFounders, shared his thoughts on the pre-seed raise in a statement saying that since work has become extremely dispersed and asynchronous, businesses need the tools to effectively communicate and develop consistency. Spoke is developing just that and when the founding team was first introduced to byFounders, everyone was astonished by their product ideas. Casper remarked that the team is thrilled to partner with Max, Jack, and Carl for what lies ahead.
