Every few weeks a news item appears that reminds us just how fast the artificial intelligence market is moving. The latest: Nous Research, a three-year-old startup, confirmed a $90 million round at a $1.5 billion valuation and at the same time announced its leap into the corporate world with AI agents for businesses, as reported by TechCrunch on October 7, 2026. Beyond the financial headline, what is interesting for anyone running an SME is something else: the sector’s focus is shifting from the «chat that answers» toward systems that execute complete tasks within real business processes.
And that is precisely the conversation we have been having with our clients for months.
What the news actually says
Summarized and without embellishment, according to the information published by TechCrunch (techcrunch.com): Nous Research develops Hermes Agent, an open-source agent that is very popular among developers, and is now launching an enterprise-oriented version that makes it possible to deploy custom agents capable of chaining together multi-step workflows while keeping data private. The round was led by Robot Ventures, with participation from Nvidia, Union Square Ventures, Menlo Ventures, Samsung and 1789 Capital.
Two details deserve attention. First: the commercial argument is not «our model is smarter», but privacy, data control and multi-step workflows. Second: the entry route has been open source, which confirms that the model layer is becoming increasingly accessible.
The underlying takeaway: if the technology becomes cheaper and standardized, the competitive advantage no longer lies in having access to the model. It lies in how you connect it with your processes and your data.
From assistant to agent: the difference that matters
An assistant answers questions. An agent executes multi-step tasks using tools: it queries a system, compares, decides according to a set of rules, writes into another system and alerts a person when something falls outside what was expected.
Let’s take a recognizable example. Receiving a supplier invoice by email, reading it, checking that it matches the order and the delivery note, detecting the discrepancy of two pallets that never arrived, recording the document and leaving the incident open for purchasing to review. That is not «generating text»: it is an administrative process with rules, with systems involved and with accounting consequences. That is where an agent delivers real, measurable value.
The practical difference for a company is that an agent needs something a chat does not require: permissions, integrations, traceability and an owner. In other words, a project, not a subscription.
AI agents for businesses: three conditions before you start
Before diving in, it is worth checking three things. They are not very glamorous, but they make the difference between a project that goes into production and a pilot that dies in a demo.
1. Organized and governed data
Without clean, accessible data, AI does not perform. If the customer master file has duplicates, if price lists live in five spreadsheets and if every branch names items in its own way, no agent is going to fix that by magic: it will only spread the disorder faster. That is why we first work on identifying the relevant data and on data governance: who accesses what, with what quality and under which regulations.
2. A narrow, measurable use case
«Let’s do something with AI» is not an objective. «We want to reduce the time it takes to record and validate supplier invoices» is, because it can be measured before and after. Our approach in AI projects is always the same: develop and run a use case tailored to the business, with a clear indicator, and scale from there. First the process that hurts, then the ambition.
3. Control: privacy, traceability and human oversight
An agent that invoices, orders materials or replies to a customer has the ability to do damage if it gets things wrong. You have to define what it can do on its own, what requires human validation and how each action is recorded. Also where the data resides and who processes it. This is not bureaucracy: it is what lets you sleep soundly and, if the need arises, audit a decision.
What is realistic today in a Spanish SME
Without inflated promises, these are the areas where we see immediate potential:
- Administration and back office. Reading and classifying documents, reconciliations, preparing recurring orders, alerts for margin deviations.
- Internal customer support. An assistant connected to the company’s actual documentation (RAG technique) that answers sales reps and technicians with verifiable information, not invented.
- Analysis and decision-making. Exploiting ERP data with Power BI and intelligent alerts: which items are turning over poorly, which customers are cooling off, which project is deviating in cost.
- Intelligent automation. Power Automate combined with AI for tasks that previously required human judgment, such as deciding which department each incident goes to.
In the sectors we work in —retail and distribution, construction, food industry, hospitality, sports centers or fleets— these use cases repeat themselves with minimal variations. And almost always the starting point is the same: the ERP.
How we put it into practice at Tisa
At Tisa we have been implementing management systems since 1987, and we are a Microsoft Partner and certified ISV in Business Central and Power Platform. That combination is what allows us to tackle AI agents for businesses without selling smoke: we know the process, we know the data and we know how to integrate it.
The usual path has three steps: consulting (identifying where the value is and what data supports it), use case (developing it, measuring it and taking it to production) and training the team so that the tool is genuinely used. If your company already works with Business Central or Dynamics NAV, the road is usually shorter than you imagine, because the data is already structured.
News like that of Nous Research confirms that the technology is advancing fast and that there will be options for every budget. The useful question is not which provider will win the race, but which process in your company deserves to be the first to be automated with good judgment.
Let’s talk about your first use case
If you are weighing up where to start, we will help you identify it and estimate its return with no commitment. Write to us at info@grupotisa.com, call us at (+34) 971 305 885 or visit grupotisa.com to learn about our custom development, Business Central and artificial intelligence services.
News source: TechCrunch, «Nous Research confirms it hit $1.5B valuation, launches AI agents for business users» (October 7, 2026).