By Alan Pelz-Sharpe
This irregular podcast series examines what is happening in the unstructured data automation market. Three topics - Thirty Minutes, that's the format!
Topics range from the state of Blockcha
5
11 ratings
The podcast currently has 45 episodes available.
The twenty-first episode of the podcast you know and love as "We Love Ugly Data!" is out. It's available in audio form everywhere you get your podcasts from and additionally in video form via YouTube (which we've embedded below). This time, the format is different, as Alan and Dan join Matt to discuss the five winners of this year's Deep Analysis Innovation Awards.
In this month's special episode:
Winner: Hyperscience
First up, Dan talks about Hyperscience and its Hypercell for GenAI, which helps organizations process large volumes of data ready for use in GenAI applications.
Winner: ai12z
Next, Matt outlines why AI12z is among this year's winners. Their approach to generative AI enables not only mid-market customers but also the systems integrators that service them to deliver AI assistants and agents.
Winner: Rossum
Dan's back to introduce Rossum's award and discuss its T-LLM (Transactional LLM) model, which it built from millions of transactional business documents.
Winner: Composable
Matt introduces Composable as the next award winner, explaining how the company focuses on organizations that have outgrown the available tooling for their AI application development and need Composable's help managing the plethora of models and inference suppliers that those heavily invested in developing GenAI applications in-house need to work with.
Winner: UiPath
Finally, Dan explains why UiPath won this year's awards for its DocPath and CommPath AI models, which were launched earlier this year.
https://www.youtube.com/watch?v=nT0RtYNIFGQ
Show notes for Series 3, Episode 11.
Innovations Awards Related Links!
Download the full Innovation Awards report.
Support the show
The twentieth episode of the podcast that you know and love as "We Love Ugly Data!" is out; available in audio form everywhere you get your podcasts from and additionally in video form via YouTube (which we've embedded below). As always, it's three topics in (just a touch over) 30 minutes, and the show notes are as usual included just below the embedded video. Matt and Alan are back again, this time discussing; the latest vendor research reports on DRUID, M-Files and Writer, how automation and AI agents are friends forever and the (really positive) state of the automation software market in Brazil.
In this month's episode:
Topic 1: New Vendor Research!
First up, Matt and Alan discuss the recently published set of vendor profiles (or whatever it is we're supposed to call them these days). DRUID is up first; an all-in-one platform for the development of conversational AI assistants and agents, which can be embedded into business processes. M-Files is a Microsoft-focused knowledge work automation that provides an innovative “no-folders” approach to document and knowledge management. Finally, Writer enables enterprises to create custom GenAI applications to assist workers in producing precise, detailed, and compliant content which can be plugged into routine tasks to help streamline processes (it's also out there raising an apparently huge funding round right now and - as Alan notes - is also an acquisition target right now too). Matt also mention Writer is one of a host of vendors now using synthetic data to speed up development new models (with all the potential issues that creates).
Topic 2: AI Agents and Automation; forever friends
Following on from last month's discussion about AI Agents in the run-up to Salesforce's Dreamforce conference, Matt's published something wrapping up how integral automation (think workflow, RPA etc) is to the success of any AI Agent exercise and the pair discuss their takes on how those efforts might play out for those looking to adopt. There's a chance to have a look at the Agentforce announcements in the rear-view mirror, reviewing the takes from Matt in his home office at the east Kent seaside and Alan on the ground in San Francisco.
Topic 3: Automation in Brazil
Back in August Alan visited Brazil to speak and catch-up with the state of the automation software market in the country (the world's 8th or 9th biggest economy, depending on whether you're using the IMF of the World Bank's numbers). Turns out that it is thriving and in the context of the other discussions about AI Agents, this puts the country in a potentially advantageous position.
https://www.youtube.com/watch?v=fkvWyu0e1s4
Show notes for Series 3, Episode 10.
Topic 1: New Vendor Research!
"New Research Dropping Today" blog post (with summaries of the three profiles).
The Deep Analysis vendor research database; links to all the vendor research we've published (you can browse as a non subscriber, but you'll need to be a subscriber to get the full reports).
The CNBC report "AI startup Writer, currently fundraising at a $1.9
Support the show
The nineteenth episode of the podcast that you know and love as "We Love Ugly Data!" is out; available everywhere you get your podcasts from and, of course, in video form via YouTube (which we've embedded below). As always, it's three topics in (just a touch over) 30 minutes, and the show notes are, as usual, included in the embedded video below. This time around, Matt and Alan are in the chairs, discussing a new research report we've just released on AI and unstructured data, how AI agents are coming and how they might be after your wallet, and the dos and don'ts of storytelling when you're presenting your software product.
In this month's episode:
Topic 1: AI & Unstructured Data, Survey Out Now
First up, Matt and Alan discuss the brand new report that Deep Analysis has collaborated on with AIIM and MFiles; "Market Momentum Index: AI and Unstructured Data Management". The report - which you can download from AIIM or MFiles directly - provides insight into the adoption and use of AI in organizations and provides information on how those same organizations manage unstructured data for AI and their view of its importance. They pick up on a couple of data points on AI adoption and how vast the sprawl of unstructured data is across IT estates (as well as mentioning that there's more related research already in the works).
Topic 2: Agents and Cash
In the second topic this month, Matt has recently posted a new blog post, "Here are the agents. They've come to collect," about a shift in how AI is likely to be paid for as the generative wave moves from assistants to agents. With Salesforce's "Dreamforce" conference only days away at the time of recording (and neatly avoiding saying anything that will break any news embargos), the pair chat about how the economics had shifted from when we first predicted a metered future for generative AI a year ago and the company's announcement of its "hard pivot" to AI agents with Agentforce. Matt also tries to extend an analogy about buffets far too far for its own good.
Topic 3: Analyst Briefings, Dos and Don'ts (Pt 2)
Over recent podcasts, Matt, Alan, and Dan discussed their dos and don'ts for analyst briefings (which we recently rounded up here, so you don't have to search through old podcasts to find them). It turned out to be quite popular, so to try and help further - and inspired by Dan's suggestion that good storytelling was a definite do - here Matt and Alan suggest their suggestions for dos and don'ts. It sadly quickly heads off-piste, and you end up with further conference recommendations (that we covered in part previously).
https://www.youtube.com/watch?v=iaU_jlchk1M
Show notes for Series 3, Episode 9.
Topic 1: AI & Unstructured Data, Survey Out Now
Our Market Mome
Support the show
The eighteenth episode of the podcast that you know and love as “We Love Ugly Data!” is out; available everywhere you get your podcasts from and, of course, in video form via YouTube (which we’ve embedded below). As always, it’s three topics in (just a touch over) 30 minutes, and the show notes are, as usual, included in the embedded video below. This time around, Matt and Alan are in the chairs, discussing 4 new Vendor Profiles (Tungsten Automation, Cognaize, Composable, and iGrafx), catching up on how the meeting of unstructured data and generative AI is working out, and – our second edition of – Analyst Briefings; Do’s and Don’ts.
In this month’s episode:
Topic 1: New Research!
We recently published the latest batch of profiles for our Vendor Analysis research (the artist formerly known as Vendor Vignettes, although we often forget that that’s supposed to be the former title and still use the phrase all the time). Here, Matt and Alan discuss the new profiles on Tungsten Automation, Cognaize, Composable, and iGrafx (a real mix of the old and the brand new here).
Topic 2: GenAI and Data: an update
Matt has just begun his third year with Deep Analysis, and to celebrate, he and Alan look at the path that generative AI and its intersection with unstructured data have taken over the last few years: from the first flush of “isn’t it amazing/scary!” to “here’s a complicated way to avoid asking it anything much” and onto the exciting world of agents and large action models.
Topic 3: Analyst Briefings, Do’s and Don’ts (Pt 2)
As analysts, we spend much of our time briefed by software vendors. Last month, Dan and Matt discussed the dos and don’ts of presenting your company and products to analysts, and now, it’s Alan’s turn to pitch in with his free advice on the subject.
Show notes for Series 3, Episode 7.
Topic 1: New Research!
Here’s the intro blog for the new vendor research: New Research.
Subscribers have access to all the vendor research, but you can browse the catalog here.
Dan’s blog post heavily features Cognaize, as discussed in the previous pod.
If you’d like access to the Composable profile, you can download it for free!
Topic 2: GenAI and Data: an update
Matt’s first blog post from (almost)2 years ago; “AI-Enabled BizApps driving the next wave of adoption.”
“The AI with a two-track mind” (AI without access to business data is a “parlor trick”).
“Models and the RAG trade” (the 5 micro ages of massive models).
Support the show
The seventeenth episode of the podcast that you know and love as "We Love Ugly Data!" is out; available everywhere you get your podcasts from and, of course, in video form via YouTube (which we've embedded below). As always, it's three topics in (just a touch over) 30 minutes, and show notes are, as usual, included below the embedded video. This time around, Matt and Dan are in the chairs, discussing how to train your LLM, The Usual Suspects (chosen by the usual analyst firms), and Analyst Briefings: Do's and Do not.
In this month's episode:
Topic 1: How to train your LLM
Inspired by a blog post of his from last year, Dan discusses the finer points of how LLMs can be trained to ensure that your IDP processes work better. He also confirms his membership in the court of King Charles, which, to be honest, wasn't necessary.
Topic 2: The Usual Suspects
Again, inspired by another blog post—this time a bit more recent—Dan wants us to remember to look beyond those in the top right-hand corner of the 2x2 diagram when we're selecting software. It's naturally focusing on IDP again, but the lesson can easily be applied across the board. Matt also suggests that we may not have paid for the associated image, but he turns out to be entirely wrong (so call off the lawyers).
Topic 3: Conference programs; what works and what doesn’t?
As analysts, we spend much of our time being briefed by software vendors. Here, Dan and Matt discuss the do's and do nots of presenting your company and products to analysts (hint, hint: go to town on talking about case studies and the actual details of projects you've done).
https://www.youtube.com/watch?v=-OmIWNC5QsA
Show notes for Series 3, Episode 7.
Topic 1: How to train your LLM
Dan's original blog post "How to Train your LLM."
Topic 2: The Usual Suspects
Dan's original blog post, "Don't Settle for the Usual Suspects."
Support the show
The sixteenth episode of the podcast that you know and love as “We Love Ugly Data!” is out; available everywhere you get your podcasts from and, of course, in video form via YouTube (which we’ve embedded below). As always, it’s three topics in (just a touch over) 30 minutes, and the show notes are, as usual, included in the embedded video below. Matt and Alan are in the chairs this time, discussing how GenAI assistants are becoming agents, how RPA is growing in the shade, and what parts of vendor conferences work (and which bits don’t).
In this month’s episode:
Topic 1: AI Assistants become Agents
For Topic 1, Matt and Alan discuss how after 2023 being the year where everyone launched an GenAI assistant, 2024 is the year when everyone is launching a GenAI agent. And to keep it confusing (or simple?) they’re all called Copilots. The pair discusses the phenomenon, based on the differences, of how the agent approach might suit some mature use cases better than others and how quickly a “decision tree” approach to developing and testing an agent will get really tricky indeed. Also, Matt references Spinal Tap (amazingly, for the first ever time on this pod).
Topic 2: Automation; zero spotlights and doing fine?
Alan’s been on the road again, most recently visiting Automation Anywhere’s “Imagine 2024” conference in Austin, Texas. In discussing all of that, the pair wonder whether being in the shade suits RPA, given the recent strong results for UiPath and general good market health (as predicted by the Task Execution market projections within the 2023 and 2024 editions of the Work Intelligence Market Analysis). Could it also be that the scramble for GenAI use cases is throwing up good opportunities for RPA that could deliver faster value?
Topic 3: Conference programs; what works and what doesn’t?
As analysts, we are privileged to attend many in-person conferences and have primo seats and often luxuries like refreshments and power sockets (I know, it’s a gilded existence). Given that experience, Matt and Alan each pick one thing that works well at these conferences and another that should be avoided at all costs. Unsurprisingly, this topic took this edition of the podcast over time.
Show notes for Series 3, Episode 4.
Topic 1: AI Assistants become Agents
Matt’s recent blog post concerns Salesforce’s strategy regarding GenAI assistants and agents.
“Copilots may have misdiagnosed the problem; humans don’t do what we thought they did.” Matt’s blog post about how we might be overestimating how much human-like experience we can gain from knowledgebases.
This short clip from the film “Spinal Tap” should explain Matt’s reference to lukewarm water.
Topic 2: Automation; zero spotlight and doing fine?
Alan’s recent blog post discusses his thoughts on having visited Automation Anywhere’s “Imagine 2024” conference.
Support the show
The fifteenth episode of the podcast that we know and love as “We Love Ugly Data!” is out; available everywhere you get your podcasts from and, of course, in video form via YouTube (which we’ve embedded below). As always, it’s three topics in (just a touch over) 30 minutes and show notes are, as usual, included below the embedded video. Matt and Alan are in the chairs this time, discussing new vendor profiles and AI readiness.
In this month’s episode:
Topic 1: 5 New Vendor Vignettes
First up, Matt and Alan discuss the latest set of – what used to be called – Vendor Vignettes, that have recently been published; debuts for ai12z, Docuvela and Reshape AI and completely updated profiles for Appian and Apromore. The pair discuss how these research updates connect to current trends and how they fit into the existing application landscape for organizations who might look to adopt them.
Topic 2: 5 Questions for AI Readiness
In conjunction with AIIM (Association for Intelligent Information Management), Deep Analysis (along with Kash Kompella of RPA2AI) have compiled an infographic that provides organizations with 5 key questions they need to be able to positively answer before they should begin any AI project (especially generative AI). Here Matt and Alan discuss the individual steps, why they are all important and also, why the document sports a cartoon Alan at the center.
Topic 3: Why are process tools having such a hard time?
Alan’s been on the road a lot in the last few weeks and – as he does – having a lot of conversations with the people he meets. From these conversations, he reckons that as much as Deep Analysis highly rate the importance of Task and Process Mining, they are proving a hard sell in the marketplace (especially Process Mining). Here the pair discuss why that might be, what approaches they’ve heard are being used to offset these challenges and what the future might look like (the real future, not whatever weird and wild guesses Gartner are making this time).
Show notes for Series 3, Episode 4.
Topic 1: 5 New Vendor Vignettes
Alan’s introductory blog post on the latest set of Vendor Vignettes.
Topic 2: 5 Questions for AI Readiness
Download the AIIM infographic “5 Questions – Is Your Organization Ready for AI?” [COMING SOON]
Alan’s recent blog posts; “Orchestrating AI – Appian” and AI Readiness – “Ready for what, exactly?”.
Topic 3: Why are process tools having such a hard time?
To join the mailing list and receive the newsletter (or challenge us as to how your attempts to popularize Task ad Process Mining are going), contact us.
Support the show
The fourteenth episode of the podcast that we know and love as “We Love Ugly Data!” is out; available everywhere you get your podcasts from and, of course, in video form via YouTube (which we’ve embedded below). As always, it’s three topics in (just a touch over) 30 minutes, and the show notes are, as usual, included in the embedded video below. Matt and Alan are in the chairs this time for a Work Intelligence-focused episode.
In this month’s episode:
Topic 1: Work Intelligence Market Analysis 2024-2029, Out Now!
We kick off this episode by discussing the newly released “Work Intelligence Market Analysis 2024-2029,” which is now out for subscribers (and for one-off purchase for nonsubscribers). Matt and Alan discuss the constituent parts that make up Work Intelligence and how they fit together before going on to look at some of the biggest revenue contributors to that market and a quick discussion about some of the new data points that have debuted in this new edition of the report (geographic location and company size).
Topic 2: Farewell, Workfellow
Very much still in the area of Work Intelligence, Matt and Alan talk about the sad demise of one of our favorite Mining Intelligence start-ups, Workfellow. Pulling some further data from the Work Intelligence report, the pair discuss the relative performance and funding in particular for Process Mining and Task Mining vendors and how even that data is somewhat skewed by the presence of one, high earning and fabulously funded player in that sub-market.
Topic 3: RAG SLAM
Matt has recently published a blog post, “Models and the RAG trade”, discussing how the proposed enterprise use of generative AI has shifted quickly from it being a magic box to how it’s necessary to build and support a lot of information scaffolding around it to produce useful results. Here, the pair discuss the differences between using RAG (Retrieval Augmented Generation) and domain-specific, smaller language models (which you could call a SLAM) and where current thinking on each is likely heading. It also touches on the news – covered in another recent blog from Matt – of the hiring of (well-funded LLM start-up) Inflection’s core team by Microsoft.
Show notes for Series 3, Episode 4.
Topic 1: Work Intelligence Market Analysis 2024-2029, Out Now!
The announcement blog post for the WI Market Analysis report update.
Details on how to purchase the report for non-subscribers.
Topic 2: Farewell, Workfellow
“Workfellow – a sad loss”, Alan’s blog post signaling farewell to one of our favorite start-ups.
Topic 3: RAG SLAM
Matt’s blog posts, “Models and the RAG trade” and “Generative AI: focus on the consumer and the consumption”.
2020 research paper from Meta/UCL that firs
Support the show
The thirteenth episode of the podcast that we know and love as “We Love Ugly Data!” is out; available everywhere you get your podcasts from and, of course, in video form via YouTube (which we’ve embedded below). As always, it’s three topics in (just a touch over) 30 minutes, and show notes are, as usual, included below the embedded video. Matt and Dan in the chairs this time, in a very IDP focused episode.
In this month’s episode:
Topic 1: Irrational Exuberance?
First, Dan provides an introduction to the IDP Market Analysis that he’s recently completed (and you can purchase copies of right now). The pair then move on to talk about one of the key findings in the report itself, which suggests that the vendors surveyed for the research are very bullish about their projected increases in short-term revenue. Dan balances this against his own market projections and Matt compares start-ups with hungry baby birds (it makes sense in context, trust us).
Topic 2: GenAI’s outsized impact
Referencing another data point from the IDP report, Dan talks about how GenAI continues to wield a huge influence in the product roadmaps for IDP vendors and points to a couple of recent specific announcements (see show notes below) to illustrate that fact. Matt chips in with a reminder of the fact that metered access to GenAI is coming over the horizon.
Topic 3: Data and Research Efficacy
Referencing a recent blog post on the subject, Matt and Dan discuss the challenges in gathering and maintaining research data for market analysis reports, ensuring that it is representative and how they go about choosing what is and what isn’t added into the final published analysis. There’s also some insight into who are the real buyers (and readers) of this sort of specialized research.
Show notes for Series 3, Episode 3.
Topic 1: Irrational Exuberance?
Announcing the 2024 IDP Market Report blog post by Dan
Purchase the report itself, here.
We Love Ugly Data! – The Deep Analysis Podcast: Series 2, Episode 5 previous pod episode where Matt and Dan discuss the history of IDP
Topic 2: GenAI’s outsized impact
More startups are using GenAI for IDP blog post by Dan
Rossum launches its own LLM blog post by Dan
Tungsten Automation (Kofax) launches GenAI Copilots for intelligent automation blog post by Dan
Topic 3: Data and Research Efficacy
Understanding scale and location: Another new lesson from Work Intelligence blog post by Matt
Support the show
The twelfth episode of the podcast that we know and love as “We Love Ugly Data!” is out; available everywhere you get your podcasts from and, of course, in video form via YouTube (which we’ve embedded below). As always, it’s three topics in (just a touch over) 30 minutes, and show notes are, as usual, included below the embedded video. Matt and Alan are in the chairs this time around.
In this month’s episode:
Topic 1: Don’t Trust in Magic
Here, the discussion centers upon Matt’s recent blog post, where he discusses that some organizations expect generative AI to find a role in their working methods without a clear idea up front as to what that might be. During a discussion about how Task and Process Mining can prevent this situation, Alan muses on Free Jazz while Matt worries about the dangers of just giving everybody a hammer to play with.
Topic 2: Business Forms & AI
Alan talks about the recent news that the business forms industry group BFMA has been folded into the information management association AIIM and how we often overlook how important forms are to how organizations work. Naturally the discussion ends up in a reminder that for many people (customer, works and partners) these forms are the entire interface that they have with important parts of an organization or even their entire experience of working with them.
Topic 3: Work & Trust
Following a blog post of Matt’s early in January about how the Horizon/Post Office scandal in the UK illustrates how mistrust of workers undermines a lot of what organizations would like to do to improve their ways of working, the pair discuss the topic and remind people that aside from anything else, if you’re in London, you should definably visit The Postal Museum and ride the Mail Rail. Alan goes on to mention situationism because, of course, he does.
Show notes for Series 3, Episode 1.
Topic 1: Don’t Trust in Magic
Matt’s blog post “Don’t Trust in Magic: A new lesson from Work Intelligence”
Ethnic Heritage Ensemble
Topic 2: Business Forms & AI
Alan’s blog post “Business Forms & AI”
The announcement of the BFMA joining AIIM
Topic 3: Work & Trust
Matt’s blog post “Following new horizons through Work Intelligence”
The London Postal Museum
Guy Dubord
Support the show
The podcast currently has 45 episodes available.