by Gary Mintchell | Jul 1, 2025 | Generative AI
Deepgram is an intriguing company. Have they solved the problem that Apple still misses with Siri or Amazon with its new Alexa? They bill themselves as “World’s Only Enterprise-Ready, Real-Time, and Cost-Effective Conversational AI API.” They have developed a voice AI platform.
In addition, its CEO Scott Stephenson has become a YouTuber with a YouTube channel (billed as a podcast, but it isn’t one), “The Scott Stephenson AI Show” — A No-Hype, Deep-Dive Podcast on the AI Revolution. Oh, he’s also on Spotify. I am not. I download podcasts on Overcast. I haven’t the time to watch many 40+ minute YouTube videos. I’ve watched much of this one. He does provide a knowledgeable overview in this episode.
Back to the Deepgram API.
Deepgram announced the general availability (GA) of its Voice Agent API, a single, unified voice-to-voice interface that gives developers full control to build context-aware voice agents that power natural, responsive conversations. Combining speech-to-text, text-to-speech, and large language model (LLM) orchestration with contextualized conversational logic into a unified architecture, the Voice Agent API gives developers the choice of using Deepgram’s fully integrated stack (leveraging industry-leading Nova-3 STT and Aura-2 TTS models) or bringing their own LLM and TTS models. It delivers the simplicity developers love and the controllability enterprises need to deploy real-time, intelligent voice agents at scale. Today, companies like Aircall, Jack in the Box, StreamIt, and OpenPhone are building voice agents with Deepgram to save costs, reduce wait times, and increase customer loyalty.
I can no longer download and play with software like in the old days. I’d suggest that if you’re a developer and need a voice assistant, try it out.
For teams taking the DIY route, the challenge isn’t just connecting models but also building and operating the entire runtime layer that makes real-time conversations work. Teams must manage live audio streaming, accurately detect when a user has finished speaking, coordinate model responses, handle mid-sentence interruptions, and maintain a natural conversational cadence. While some platforms offer partial orchestration features, most APIs do not provide a fully integrated runtime. As a result, developers are often left to manage streaming, session state, and coordination logic across fragmented services, which adds complexity and delays time to production.
Deepgram’s Voice Agent API removes this burden by providing a single, unified API that integrates speech-to-text, LLM reasoning, and text-to-speech with built-in support for real-time conversational dynamics. Capabilities such as barge-in handling and turn-taking prediction are model-driven and managed natively within the platform. This eliminates the need to stitch together multiple vendors or maintain custom orchestration, enabling faster prototyping, reduced complexity, and more time focused on building high-quality experiences.
In addition to the Voice Agent API, organizations seeking broader integrations can leverage Deepgram’s extensive partner ecosystem, including Kore.ai, OneReach.ai, Twilio and others, to access comprehensive conversational AI solutions and services powered by Deepgram APIs.
Key capabilities include:
- Flexible Deployment: Run the complete voice stack in cloud, VPC, or on-prem environments to meet enterprise requirements for security, compliance, and performance.
- Runtime-Level Orchestration: Deepgram’s runtime supports mid-session control, real-time prompt updates, model switching, and event-driven signaling to adapt agent behavior dynamically.
- Bring-Your-Own Models: Teams can integrate their own LLMs or TTS systems while retaining Deepgram’s orchestration, streaming pipeline, and real-time responsiveness.
In addition to control and performance, the Voice Agent API is built for cost efficiency across large-scale deployments. When teams run entirely on Deepgram’s vertically integrated stack, pricing is fully consolidated at a flat rate of $4.50 per hour. This provides predictable, all-in-one billing that simplifies planning and scales with usage.
by Gary Mintchell | Jun 25, 2025 | Robots, Sensors, Software, Technology
This is part two of my reports from the Hexagon Live Global Event. I had been to one previous event for only a day. Hexagon is such a large company comprised of many parts that I had a bit of struggle understanding it all.
The core Hexagon involves measurement, metrology. They have precision measurement tools for the small, medium, and very large targets. Tools for building applications complement these instruments.
Hexagon also comprises much software, having acquired Intergraph years ago and others since. Company focus has become easier with the announced spinoff of much of the software business into a new company called Octave.
Octave
Mattias Stenberg, who is leading Octave, explained the new company’s focus taking the musical analogy of an octave—taking it to the next level. The core of the new company consists of these four businesses from Hexagon:
- Asset Lifecycle Intelligence—Intelligence that drives decision-making efficiency and lifecycle value creation
- SIG—Safety, Infrastructure, and Geospatial—Act on information to save lives, improve infrastructure and enhance services
- ETQ—ETQ Reliance—ETQ Reliance is a cloud-native quality management system solution (QMS), powered by an agile platform that drives 40+ best-in-class applications adaptable to your unique environment. ETQ is the leading provider of quality, EHS and compliance management SaaS software, trusted by the world’s strongest brands.
- Bricsys—Hexagon AB, a global leader in digital solutions, today announced the acquisition of Bricsys, a fast-growing developer of CAD (computer-aided design) software that has been at the forefront of providing open, collaborative construction technology solutions since its founding in 2002. Its CAD platform, BricsCAD, supports 2D/3D general, mechanical, and sheet metal design and building information modelling (BIM) in one system.
I don’t know if this was supposed to be another musical reference, but as a guitarist, I’ll take it as such. Octave is Intelligence at Scale. They see themselves not just as a builder of software, but as helping customers evolve, adapt, predict, prevent by providing pre-trained agents. The platforms will be embedded, context aware, mission critical systems, validated.
This comment struck me. I concluded my first report with the thought that software becomes more powerful, yet it’s still trying to solve the problems I had in 1977. With power comes complexity. Stenberg noted a survey they conducted with C-level executives see more software, more complexity, more dashboards…and yet, less actual visibility. Systems that don’t talk to each other creating silos.
My concluding thought on my last essay was that we must not have a technical problem—we must have a people problem.
Only 20% of execs told them they are getting something from digital transformation. OK, I can’t resist thinking of an irony—yet they order their employees to use AI, or else???
They envision a process where customers build digital first, where the digital twin is a reality (he calls it “mirrorworld”). These will enable the movement from reactive to predictive (another future vision I’ve written about for 20 years or more).
A final vision—Create self-aware, resilient infrastructure.
Robotics
Moving on to another product line—robotics. I shot a short video of Hexagon’s newest robot—Aeon. This “humanoid”, or human-form-factor, robot exhibits quite advanced ability to do the work of human assemblers. When I asked why they developed the robot to look and act like a robot, they told me that existing work stations are designed for humans. Therefore, this is an easy replacement for non-existent human workers on the assembly line.
Digital Twins
Returning to digital twins. I spoke with Jeremy Treverrow about uses of digital twins. Initially, customers could use Hexagon’s precision measurement technology to create a digital twin of a component part. Perhaps this is a service and repair part no longer in production with perhaps no good design information existing.
Using the Hexagon Design X software set, the imported digital twin can be exported in an igis file, used for simulation, and can even design a manufacturing process around it.
A lot of power.
by Gary Mintchell | Jun 21, 2025 | Data Management, Software
I recently wrote an article for my website about technology complexity within industrial technology. Engineering managers have stood at conferences pleading with the standards and technology developers to find ways to simplify interfaces and connectivity.
OPC Foundation keeps adding layers of companion specifications. ODVA members listened to engineers who need help implementing EtherNet/IP (or just ethernet networks) and proceeded to ignore the plea. Paul Miller, an analyst at Forrester reported from a survey where 90% of executives reported data problems from their digital transformation. 71% reported measurement related data problems.
Mattias Stenberg, head of the new software company spinning out from Hexagon called Octave, reported from another survey his group has performed that only one in five executives thought they were getting any value from digital transformation.
The Vice President of Product Development of the company where I worked in the 1970s (back in the day before layers of vice presidents) offered a job to me to leave manufacturing and become his data manager. He was prescient. 45 years later, companies are still trying to manage data. Solutions have become more complex, technology has advance exponentially, yet we still have problems gathering, refining, contextualizing, and using data.
These thoughts were generated from the Hexagon Live Global Conference I attended this week in Las Vegas. I have a lot of trouble wrapping my head around just who Hexagon is. Evidently, I’m not alone. But the company is making it easier by splitting off four groups into Octave.
The simplest definition, yet also most definitive, came from Ola Rollén Hexagon Chairman recounting the company’s 25-year history of growth. “Hexagon is the world’s most sophisticated measuring tape.” Indeed, several of my interviews delved into the world of accurately measuring the very large and the very small. This year’s slogan, “When it has to be done right.”
The new ATS800 laser tracker can easily capture complex shapes with up to micron precision. The company released Autonomous Metrology Suite, software developed on its cloud-based Nexus platform that is designed to transform quality control across manufacturing industries worldwide. By removing all coding from coordinate measuring machine (CMM) workflows, it helps manufacturers speed up critical R&D and manufacturing processes as experienced metrologists become harder to find.
Hexagon and several partners are solving what has been an intractable and troubling problem—data locked into paper-based formats such as pdf files. Several demonstrated the ability to read text and pdf documents that are unstructured data, use a form of AI to tag the data, and then extract to a useable database. This is truly a great advance. Several workforce solutions designed to give companies the ability to attract younger workers into technical positions were demonstrated on the show floor.
Stenberg talked of another problem executives cited—data silos that prevent people from using data to make good decisions. I have been writing about solutions designed to break through data silos for 25 years. I’m beginning to wonder if it is not a technology problem. Perhaps it’s a people problem.
by Gary Mintchell | Jun 19, 2025 | Robots, Software
Everyone wants to be NVIDIA’s best friend. This friendship focuses on the world of robotics and simulation. In brief:
Flexiv launches the Flexiv-Isaac Bridge App, empowering developers to design, test, and deploy force-controlled robotics applications in hyper-realistic virtual environments.
The interesting thing is the force-controlled part. I remember an application I wrote about a few years ago where the developer of a robotic prosthetic arm had a design goal of being able to pick up a grape with squishing it. I watch these technologies closely anticipating even greater use cases that will help us all.
Flexiv announced release of the Flexiv-Isaac Bridge App, bringing high-fidelity force-control simulation to NVIDIA’s Isaac Sim. This partnership enables robotics developers and end-users to program, model, test, and deploy complex force-controlled, AI-empowered robotics applications in simulated environments that closely mimic contact-rich real-world conditions.
To highlight this new capability, Flexiv’s engineering team released a video in which a simulated Rizon 4 robot completed the classic Tower of Hanoi puzzle in Isaac Sim. The simulation exactly replicated the robot’s real-world movements and showcased its force-controlled “hole search” and compliant movement capabilities. This underscores Flexiv’s commitment to minimizing the sim-to-real gap to improve training, programming, and operational performance. Additionally, this demonstration emphasizes Flexiv’s drive to ensure seamless compatibility with one of the world’s most widely used virtual robotics platforms.
Isaac Sim enables developers to build hyper-realistic, detailed virtual environments, while Flexiv’s Elements programming system allows robotic applications to be effortlessly programmed and refined. With the Bridge App connecting these tools, customers are empowered to create applications, build digital twins of their facilities, design mission profiles, and run high-fidelity virtual tests before deploying robots in real-world scenarios. These simulations provide valuable insights into application performance, risk assessment, and operational efficiency.
By leveraging both Isaac Sim’s ability to generate real-world simulations and Flexiv Elements’ support for simulating real-world force-based actions, Flexiv aims to accelerate the application development cycle. With developers now able to refine robot movements and iteratively test applications from anywhere in the world, development costs can be significantly reduced, while remote support can ensure greater reliability in real-world deployments.
In its continued commitment to community-driven innovation, Flexiv has made its Tower of Hanoi codebase freely available on GitHub. This initiative encourages developers, academics, and customers to build upon Flexiv’s work, fostering a collaborative ecosystem that promotes creativity and customization.
By combining force-controlled robotics and effortless programming with NVIDIA’s cutting-edge simulation tools, Flexiv is revolutionizing development workflows. This leads the way toward safer, smarter, and more adaptable robotic systems that transform both application development and human-robot interaction.
by Gary Mintchell | Jun 18, 2025 | Security, Services
Speaking of Honeywell from yesterday’s post, here is another release, this one from their User Group meeting that are, of course, announcing AI use cases. They bring in another buzz word from the automation market—autonomy.
Announcements include:
AI-enabled cybersecurity solutions—Honeywell Cyber Proactive Defense and Honeywell OT Security Operations Center.
Expansion of the Honeywell Digital Prime platform to encompass an enterprise-wide set of solutions that effectively test and modify engineering projects before implementation.
Some details:
- Honeywell Cyber Proactive Defense, which is designed to enhance cybersecurity for industrial environments by proactively identifying and mitigating potential cyber threats before they manifest into attacks. By utilizing AI and behavioral-based analytics, the solution helps detect anomalies in OT cyber behavior by establishing a comprehensive baseline of system operations and then provides actionable insights designed to strengthen OT cyber defenses. The software also features deception technology, which uses decoys within the network to help divert attackers from valuable assets.
- Honeywell OT Security Operations Center, a vendor-agnostic and agentless service designed to provide industrials with advanced capabilities tailored to OT environments to monitor for early signs of a cyberattack. The offering integrates on-site incident management services, providing a 24/7/365 holistic view of the cyber threat landscape for users.
- Honeywell Digital Prime Ecosystem, which now features three core Honeywell offerings – Solution Enhancement Support Program (SESP), Enabled Services and Assurance 360 – in one platform. Through consolidation, users can now leverage deep domain knowledge to optimize control systems and improve maintenance and operational effectiveness across an entire organization. It will also offer near real-time performance insights that can help users achieve desired outcomes more quickly, while requiring less reliance on the technical expertise of an experienced workforce.
Honeywell Cyber Proactive Defense and OT Security Operations Center are now available globally. The expanded version of the Honeywell Digital Prime ecosystem will be available to customers in Q4 2025.
by Gary Mintchell | Jun 10, 2025 | Manufacturing IT
Probably 90% of the press releases I have been receiving manage to work AI into the story. Most refer not to the Large Language Models that have captured the imagination of the tech press. The most useful are machine learning (ML) and some pattern matching that we’ve had for decades.
That’s OK. Just read through the hype focusing only on the application. Is it real? Is it useful? Does it solve a customer need? This is a story from Siemens about a customer actually using one of its “AI” products.
- Senseye Predictive Maintenance significantly improves machine maintenance at Sachsenmilch with AI-powered prediction algorithms
- Early detection of the end of service life of a pump; pilot project pays off
- More automation is planned by integrating Senseye Predictive Maintenance with SAP Plant Maintenance (SAP PM)
Siemens has been supporting Sachsenmilch Leppersdorf GmbH milk processing plant in Germany on its path toward developing a predictive maintenance system. Siemens’ AI-powered solution, Senseye Predictive Maintenance, helps Sachsenmilch ensure continuous operation 365 days a year following strict quality standards.
Sachsenmilch produces a variety of products from milk, butter, yogurt, cheese, and dairy derivatives for baby food to bioethanol in its state-of-the-art and almost fully automated facilities. Every day 4.7 million liters of fresh milk are delivered for processing, the equivalent of 170 truckloads. It’s essential for the company’s equipment to operate 24/7 and for the production facilities to be nearly 100 percent available.
The production environment at Sachsenmilch in Leppersdorf features modern interconnected machines that generate large volumes of data – an ideal setting for a pilot project using Senseye Predictive Maintenance, the advanced predictive maintenance solution.
Senseye Predictive Maintenance utilizes AI algorithms to identify both immediate and future machine issues, which allows proactive maintenance to be performed and prevents downtime. This capability has proven to be extremely valuable in Sachsenmilch’s heterogeneous production environment during the pilot project.
One of the biggest challenges was analyzing relevant plant data like temperature, vibration levels, and frequencies to detect anomalies early on and draw the right conclusions. The implementation process involved a careful analysis of specific failure scenarios and the integration of existing data from the control system. New vibration sensors and the Siplus CMS 1200 measurement system for vibration monitoring were also installed.
Siemens supported the maintenance team at Sachsenmilch with technical and project management expertise. “What we like about this project is that Siemens has know-how on both the technological and the technical sides as well as in project management,” said Roland Ziepel, Technical Manager and head of project management at Sachsenmilch in Leppersdorf. After being trained and the solution’s implementation, the Sachsenmilch team was able to independently continue and successfully complete the pilot.
The pilot with Senseye Predictive Maintenance has already achieved significant cost savings by reducing unplanned downtime. “We can confirm that the pilot project with Senseye Predictive Maintenance has already paid off. Detecting a faulty pump at an early stage saved us a lot of expense – in the low six figures,” Ziepel concluded.
Building on this success, Sachsenmilch plans to further integrate Senseye Predictive Maintenance with their SAP Plant Maintenance System, with the goal of automatically transferring maintenance notifications from the Siemens solution to SAP Plant Maintenance to improve maintenance planning.
In addition, recommendations for data-driven maintenance provided by the Maintenance Copilot Senseye should also be increasingly utilized to help maintenance teams with their work. This is one of the ways that Siemens supports its customers in their innovative and integrated approach to maintenance in order to ensure their long-term operational success.