Transcript
Transcript: Tech Demo by Publicus
[00:00:08 Text appears on screen: Tech Demo by Publicus.]
[00:00:12 The screen fades to Taki Sarantakis.]
Taki Sarantakis (President, Canada School of Public Service): Welcome. I'm Taki Sarantakis, President of the Canada School of Public Service. And today, we are doing our first in our new series Tech Demo Week (CSPS Tech Demo Series) from the CSPS, and Tech Demo Week is something we've done a few times. I think this is our third iteration, and what we do is we highlight Canadian technology from Canadian companies that can help public servants do their job better, faster, and serve Canadians in a more cost-efficient way.
[00:00:49 Joe Noss appears on screen.]
Today, we have Joe Noss who is the CEO of Publicus, a little company… where are you based, Joe?
Joe Noss (Founder and CEO, Publicus): Toronto.
Taki Sarantakis: In Toronto. Tell us a little bit about yourself and tell us a little bit about your company before we see your technology.
Joe Noss: Yeah, sounds good. So, I used to work at Deloitte as a government consultant and dealt with a lot of larger digital transformations. And I would say throughout my life, I've always been, first and foremost, a government nerd, and I quit my job with a belief that there's so much data that government puts out, both structured and unstructured, right? So, structured are ERP outputs or tables, reports in an Excel spreadsheet. Structured is the actual RFP document, for example, or a contract or any PDF, and all of this information is finally able to be used because of A.I. We can process it. We can automate work we do today. We can deal with a lot of the tedious tasks we have. But first, you need to clean that data up. And so, that's really what Publicus does.
And we've started on the problem of procurement where today we help Canadian businesses navigate the procurement system by gathering all this data, and what we discovered was the system we had built to process all of this procurement data was actually really good at just processing government data generally. And so, we ended up working with PSPC on a project to provide some analytics on procurement. And now, through Innovative Solutions Canada, we're working with the Digital Services branch at Procurement Canada to clean up a lot of kind of the gaps in data there today, which I can dive more into later on.
Taki Sarantakis: Yeah. So, today, what we're going to see is a specific example of data use. But if I heard you properly, it's an example of data use. It's not like you're locked into procurement. You can actually do this with ATIPs, you can do it with reports from Transport Canada, you can do this with reports from Natural Resources Canada, and on and on and on. So, the data, particularly in this instance, is from procurement but kind of data is data. Is that right?
Joe Noss: Yeah, I think that's completely accurate. And when you look at government technology generally, I think there's a few archetypes you often see, right? Case management solutions, which are meant to track decision-making or a process. There's ERP systems which are meant to track financials or units of measurement like payroll. And then, there's often document storage, right? So, SharePoint, OneDrive, places where we're keeping these PDFs, and all of these systems are really built about record keeping and compliance but they also, when brought together, allow for such insights into these processes. So, that's what Publicus is all about, right? Taking these disparate sources, bringing them together, and allowing us to actually build reports, dashboards, agentic workflows, whatever it may be, to solve government's problems in that use case.
Taki Sarantakis: Yeah. So, take a moment, punch it up, but I want our audience to really internalize that, which is we have a lot of data in the Government of Canada, terabytes and petabytes and terabytes and petabytes and on and on and on. In some ways, we're drowning in data and one of the things that we have to shift to now is taking data and gleaming insights from the data rather than being overwhelmed with the data. So, how are you? Do you got it popped up yet?
Joe Noss: Yeah, yeah, I can share screen on this right now.
[00:04:33 A webpage is shown with the text:
"Connect Your Data Sources"
"Select the systems to analyze. We unify data across ERPs, document stores, case management, and CRM"
"Proactive Disclosure Data – 500K contracts"
"Award Notice Data – Contract awards"
"Contract Documents – 3,585 pages"
"Case Management – Case records"
"CRM (Salesforce) – Contacts & accounts".]
Taki Sarantakis: There we go. Now, I want you to tell us what we are seeing. So, we are obviously seeing a platform of some type. Tell us a little bit about the platform.
Joe Noss: Yeah. So, I'm going to take you through what the process would usually look like when working with us or in a use case generally where you would define what the data sources you want are.
[00:05:01 "Proactive Disclosure Data", "Award Notice Data", and "Contract Documents" are selected.]
Joe Noss: And in this context, we're using specifically proactive disclosure data and award notice data as well as some physical contract documents, but this onboarding would usually take a few weeks. You'd define the data sources you want to include, we'd process them, make sure that we would have 99.99% accuracy, make sure that we'd have human in the loop, reviewing every kind of A.I. output to build confidence in the system before you ever start to use it.
[00:05:34 The "Continue" button is clicked on and a new page is shown.]
Joe Noss: And once those data sources are onboarded and ready to go, you can really start…
Taki Sarantakis: Now, before… go back, before you go back.
Joe Noss: Yeah.
Taki Sarantakis: So, the first two things.
[00:05:44 The initial webpage is returned to.]
Joe Noss: Sorry.
Taki Sarantakis: Sorry. The first two things that you had clicked, you'd clicked three things, the first two were public. Is the third public as well or is that…?
Joe Noss: Yeah, in this example, this is all public information.
[00:06:02 "Contract Documents" is selected.]
Joe Noss: These are contract documents that were ATIP'd essentially, right? So, this is a demonstration of bringing together ERP data, which is structured, as well as some unstructured contract documents.
[00:06:12 "Proactive Disclosure Data" and "Award Notice Data" are selected.]
Joe Noss: But it's all in the public domain.
Taki Sarantakis: Perfect. So, these are things that anybody can go. The first two anyways, they're already out there and you've just kind of pointed your platform at them. And then, the third is things that have been released through ATIP.
Joe Noss: Yeah, exactly.
Taki Sarantakis: Terrific. Okay, let's see the next stage.
[00:06:36 The "Continue" button is clicked and a new page appears with the text:
"I've connected to 3 data sources with 500K+ contracts across all federal departments and 3,585 pages of contract documents. What would you like to analyze? Pick a focus area or describe your own:
Spending by industry & department
Top vendors & concentration analysis
RFP pipeline & upcoming opportunities
Year-over-year growth trends
Contract amendment patterns
Technology stack breakdown
Or describe a custom analysis".]
Joe Noss: Yeah, and then, you'd just be able to work with our system and define what you're actually interested in.
[00:06:43 "Year-over-year growth trends", "Spending by industry & department", and "Top vendors & concentration analysis" are selected.]
Joe Noss: Be it year-over-year growth trends or spending by industry and department, the top vendors, defining ultimately what you want this dashboard or report to focus on.
[00:06:53 The "Build Dashboard" button is clicked and a loading page appears with the text:
Building your analytics dashboard
Connecting data sources
Computing aggregates
Generating visualizations
Dashboard ready.]
Joe Noss: And finally, our system will bring together all of these different inputs and generate whatever is the output you're kind of looking for.
[00:06:59 A new page appears.]
Taki Sarantakis: Now, you're really smart and you're really quick because you're from the private sector. I'm from the public sector. So, slow down and go back again. I want to see what you clicked on.
Joe Noss: Yeah.
[00:07:14 The initial webpage is returned to.]
Taki Sarantakis: So, we have three data sets as we talked about.
[00:07:17 "Proactive Disclosure Data", "Award Notice Data", and "Contract Documents" are selected.]
Taki Sarantakis: Now, continue.
[00:07:22 The "Continue" button is clicked and a new page appears with the text:
"I've connected to 3 data sources with 500K+ contracts across all federal departments and 3,585 pages of contract documents. What would you like to analyze? Pick a focus area or describe your own:
Spending by industry & department
Top vendors & concentration analysis
RFP pipeline & upcoming opportunities
Year-over-year growth trends
Contract amendment patterns
Technology stack breakdown
Or describe a custom analysis".]
Taki Sarantakis: So, now, let's look. We've got some text here. It says, I've connected three data sources, 500,000 contracts. I assume that's all automated.
Joe Noss: Yeah, and because we've already defined all of these data sources, their fields, once that's all in place, it's very easy to now use A.I. to kind of create a dashboard, a report, a workflow, whatever it may be.
[00:07:50 "Spending by industry & department", "Top vendors & concentration analysis", and "Year-over-year growth trends" are selected.]
Joe Noss: And in this context, for example, understanding spending by industry or top vendors and year-over-year growth trends, these are the types of things we can specifically say we'd like to focus on and build a dashboard.
Taki Sarantakis: All right. So, let's un-click the year-over-year growth trends.
[00:08:04 "Year-over-year growth trends" is de-selected.]
Taki Sarantakis: And now, we've clicked spending by industry and department and top vendors.
[00:08:17 The "Build Dashboard" button is clicked and a loading page appears with the text:
Building your analytics dashboard
Connecting data sources
Computing aggregates
Generating visualizations
Dashboard ready.]
Taki Sarantakis: So, I assume if we click "Build Dashboard, what we're going to get is the biggest contracts in the Government of Canada. Is that correct?
[00:08:22 A new page appears with the text:
"Procurement Intelligence – Cross-system analytics across 98 departments"
"Total Spending – $254.4B – 502,708 unique procurements"
"Canadian Owned – 49.9% – $126.9B in value"
"Unique Vendors – 17,311 – 11,260 Canadian, 6,053 Foreign"
"Unique Procurements – 502,708 – Avg. $506K".]
Joe Noss: Yeah.
[00:08:25 Joe Noss scrolls through the page, which shows a bar graph of spending over time between 2015 and 2025 for Canadian and foreign entities followed by a ranked list of top vendors.]
Joe Noss: So, we'll be able to see now the total spending, year-over-year, within the Government of Canada. We can see the top vendors.
[00:08:38 A drop-down menu at the top of the page is selected, listing various vendor industries.]
Joe Noss: We can actually filter by, for example, a specific category like IT services and consulting.
[00:08:43 "IT Services & Consulting" is selected from the drop-down menu and Joe Noss scrolls through the page, which shows an updated bar graph of spending over time between 2015 and 2025 for Canadian and foreign entities followed by an updated ranked list of top vendors, then followed by a ranked list of top buying departments and a ranked list of top contracts.]
Joe Noss: And see how that impacts the top vendors or the spending over time. We can see who those top vendors are working with, their name variations within the data today, why they're classified within our data as a foreign government. We can see the top departments they work with, their top contracts, their competitors. We can even see recent contracts, the procurement officer involved, and if you should reach out to them essentially.
Taki Sarantakis: Yeah, let's go back because part of the demo is being able to kind of drive if you're not the owner of the company, if you're not within the company. So, I imagine DND does our biggest procurements. All right. So, let's just focus on one department. Let's go DND.
[00:09:40 A drop-down menu at the top of the page is selected, listing various departments.]
Joe Noss: Where is DND?
[00:09:54 "National Defence" is selected from the drop-down menu and the page is updated with the text:
"Total Spending – $107.5B – 125,166 unique procurements"
"Canadian Owned – 44.8% – $48.1B in value"
"Unique Vendors – 10 – 3 Canadian, 7 Foreign"
"Unique Procurements – 125,166 – Avg. $859K".]
Taki Sarantakis: There we go. So, let's pause. So, what are we looking at here? $107 billion.
[00:10:03 Joe Noss scrolls through the page, which shows an updated bar graph of spending over time between 2015 and 2025 for Canadian and foreign entities.]
Taki Sarantakis: Is this procurement last year, year before?
Joe Noss: This is over the past ten years.
Taki Sarantakis: All right, and can we make it the last five years?
Joe Noss: Yes. Right now, we can't change the timeframe but we can just look specifically.
Taki Sarantakis: All right. So, let's see who the biggest contractors were over the last five years cumulatively?
[00:10:28 An updated ranked list of top vendors is shown:
1. Seaspan (Foreign) – 18 aliases – 66 contracts – Aerospace, Defence & Marine – $13.7B
2. Skyalyne Canada Limited (Canadian) – 1 contract – Aerospace, Defence & Marine – $11.2B
3. Halifax Shipyard (Canadian) – 9 aliases – 24 contracts – Aerospace, Defence & Marine – $9.3B
4. PCL Construction (Canadian) – 11 aliases – 106 contracts – Architecture, Engineering & Construction – $5.5B
5. CAE (Canadian) – 27 aliases – 114 contracts – Aerospace, Defence & Marine – $5.4B
6. F-35 Lightning II Joint Program (Foreign) – 1 contract – Aerospace, Defence & Marine – $4.3B
7. Airbus Defence and Space (Foreign) – 4 aliases – 208 contracts – Aerospace, Defence & Marine – $3.6B
8. Bell Canada (Canadian) – 8 aliases – 1041 contracts – Telecommunications & Media – $3.6B.]
Joe Noss: Yes, this is cumulative over the past ten years.
Taki Sarantakis: All right, so Canadian, foreign, Canadian, Canadian, Canadian, CAE. Can we now click on one of those? Let's look at the F-35. Can we get more insight on the F-35?
[00:10:44 "CAE" is selected.]
Joe Noss: Yeah, yeah, I just clicked on CAE. And I'd also specify here, Taki, that when we say foreign and Canadian, this is based on ultimate beneficial ownership. As a part of this analysis, you can define whatever way you'd want to classify organizations. That's part of the enrichment we can do, and we can instead say, let's use the country label, which is what we use today within the Government of Canada's analysis, right?
[00:11:25 "F-35 Lightning II Joint Program" is selected.]
Taki Sarantakis: So, that's really important because what you're saying is, if ISED comes up with a new definition of Canadian, you've already got the data. You just filter differently.
[00:11:37 Joe Noss scrolls down the page, which shows the text:
"F-35 Lightning II Joint Program Office (JPO) – Foreign"
"Ownership classification: The research explicitly identifies the F-35 Lightning II Joint Program (JPO) as a United States Department of Defence (DoD) government organization, not a private company, and is headquartered in Arlington, Virginia"
"Confidence: High"
"Total Value – $4.3B"
"Contracts – 1"
"Avg Contract – $4.3B"
"Industry – Aerospace, Defence & Marine"
"Top Buying Departments – 1. National Defence – 1 contracts – $4.3B"
"Top Contracts – 1. Aircraft Parts – National Defence – 12/21/2022 – $4.3B"
"Top Competitors: Airbus Defence and Space (Foreign) – $3.8B – 1 shared departments – 208 contracts, CAE (CA) – $5.4B – 1 shared departments – 114 contracts, Bell Textron Canada Limited (Foreign) – $3.1B – 1 shared departments – 568 contracts"
"Their last completed project was "Aircraft parts" with the Technology Modernization Team at National Defence (2022)"
"Reach out to Patricia Moore to learn more about their experience working with F-35 Lightning II Joint Program (JPO)".]
Taki Sarantakis: If the next IP strategy… or if there are different definitions of Canadian, depending on whether it's military or whether it is non-military, the data is already there and you just change filters. Is that correct?
Joe Noss: Exactly, and these are the types of things which you just define with us and we're able to kind of implement accordingly.
Taki Sarantakis: All right. So, we've got total value of 4.3 billion you see. Now, could this tell me or tell an interested person…
[00:12:13 A textbox pops up with additional contract details.]
Taki Sarantakis: Perfect. So, will it tell me how much has actually been spent or we don't know that yet?
Joe Noss: This is coming from proactive disclosure, so it should be the payment being logged essentially against the program.
Taki Sarantakis: Right. So, basically, if you are outside the government and you want this information, it's all the information that's public, which is quite significant. But if you were a user within a department, if you kind of pointed that at your intranet data, you could add even more data to this.
Joe Noss: 100%, yeah.
Taki Sarantakis: All right. So, let's see, let's look at the F-35. So, we've got competitors, Airbus, CAE, Bell Textron. Let's scroll down, their latest air parts, National Defence. Let's look at all contracts. So, it's one big contract?
Joe Noss: Yeah.
Taki Sarantakis: All right. Let's see what we can see from that. You had like a little window where you popped up.
[00:13:30 The textbox with additional contract details pops up with the text:
"Reference Number – C-2023-2024-Q2-00001"
"Procurement ID – W847A-180210"
"Description – Aircraft parts"
"Canonical Vendor – F-35 Lightning II Joint Program Office (JPO)"
"Original Vendor Name – F-35 Lightning II Joint Program Office (JPO)"
"Ownership Status – Foreign"
"Country of Vendor – United States of America".]
Taki Sarantakis: Perfect. So, what are we looking at here?
Joe Noss: Yeah. So, here, you can actually see the contract details based on the proactive disclosure data. You're able to see the vendor name. You can see our normalized vendor name, right? Which is important. For some of the proactive disclosure data and data generally within government, there's hundreds, if not thousands, of variations for a company like Deloitte, right? So, our system is very good at identifying all these variations, normalizing them, and bringing it together, and that's a part of the challenge that we often see with government data, is really standardizing it across different formats which is leg work that no one really wants to do and is the type of thing that A.I.'s really good at.
Taki Sarantakis: Right. So, if I understand you properly, this is the, and I'm going to pronounce this wrong, canonical vendor, is kind of the beneficial ownership group. And when we say IBM, it captures IBM Canada, IBM Ottawa, IBM 794 Ontario Ltd. It kind of aggregates all of the subcontracting or the alternate names into who is kind of the beneficial owner. Is that correct?
Joe Noss: 100%, yeah. And even for Deloitte, there's Deloitte LLP, there's Deloitte Legal, right? There's all these different entities, and it brings it together.
Taki Sarantakis: So, you can see the power of getting insight from data and that is remarkably powerful.
[00:15:09 A list of contracts is shown with reference number, description, department, value, and ownership information for each.]
Taki Sarantakis: I can tell you, Joe, within the Government of Canada, a lot of this is very manual. A lot of this is e-mails back and forth between departments, units within departments, reconciling kind of what's in your system versus what's not in your system yet. Sometimes, the system doesn't catch that IBM is the same as IBM Ontario Ltd. which is also the same as IBM 794 Ltd. Tell us a little bit about the categories. So, I see here you have IT services and consulting.
Joe Noss: Yeah.
Taki Sarantakis: Let's see what other categories you've kind of templated.
[00:16:01 A drop-down menu at the top of the page is selected, listing various vendor industries.]
Joe Noss: Yeah. So, this is just based on the vendor industry, right? So, software and cybersecurity, staffing and professional services. This was based on simplicity for the analysis we were doing. But ultimately, we can go as deep or as wide or whatever classification system you want. For example, on CanadaBuys today, they use the UN's procurement classification code. We could use that to classify data. We could use the NAICS code, right? The whole point here is any type of enrichment you want on this data, be it cleaning up vendor names, be it classifying things a certain way, that's the type of enrichment we're able to do with our agents.
[00:16:43 A chatbot page pops up with the text:
"Welcome. I have access to 502,708 contracts ($254B total spend) across 98 departments, plus 3,585 pages of contract documents. I can help you explore vendor spending, ownership classification, department breakdowns, and contract details. What would you like to look at?
Which IT vendors work with more than 10 departments?
What departments work with Microsoft and what's the total spend?
Show me VMware exposure across government
Show me the Dalian-Coradix task authorizations".]
Joe Noss: And I'd add you're actually just able to chat with this data, right? So, we could ask questions like, okay, which IT vendors work with more than ten departments?
[00:16:51 "Which IT vendors work with more than 10 departments?" is selected and a lengthy response is auto-generated by the chatbot.]
Joe Noss: And the data will actually be cited exactly where we're pulling this information from. There'll be complete auditability into any transformation that's been made, and I know with the Government of Canada and governments generally, that's extremely important. You need to know, what are the document sources that are kind of the input to this analysis, what are the contracts that are being cited, and in this context, the system specially designed exactly for that, right? With the requirements of the Government of Canada in mind.
Taki Sarantakis: So, let me ask it a question since it has conversational capacity.
Joe Noss: Yeah.
Taki Sarantakis: Give me the name of the three companies who have the most amendments in their contracts?
[00:17:45 "Who are the three companies who have the most" is typed into the chat box and then backspaced.]
Joe Noss: Yeah. So, I don't have amendment data in this.
Taki Sarantakis: Okay.
Joe Noss: But that's a good one, yeah.
Taki Sarantakis: What data do you have where I could ask like the three…
Joe Noss: Top vendors with any department, you could do top vendors in any category, top departments in any category, years in terms of spending, yeah.
Taki Sarantakis: Who are the top three companies in cybersecurity contracts with the Government of Canada?
[00:18:23 "Who are the top three companies in cybersecurity contracts with the Government of Canada?" is typed into the chat box and a lengthy response is auto-generated by the chatbot.]
Joe Noss: And with these types of questions, we could automate this report so you get it bi-weekly to your e-mail to keep track of this specific question, so cybersecurity. We could actually just look to generate a report from here.
[00:19:00 "Show me VMware contracts expiring in 2025" is selected and a lengthy response is auto-generated by the chatbot.]
Joe Noss: I'll use VMware as another example, where there was recently a breach in September, I believe, where understanding which departments have contracts with VMware is critical in order to quickly deal with the issue, right? Without this kind of visibility, you can't do that easily.
[00:19:19 "Generate a report from this analysis" is selected and a report is auto-generated by the chatbot.]
Joe Noss: And then, you can just generate a report. You can ask it to create a PDF that you can use and share with your team.
[00:19:32 "Agent Builder" at the top of the screen is selected and a textbox pops up with the text:
"Select the data sources this agent will monitor:
Proactive Disclosure – 500K+ contracts
Award Notices – Contract awards
Contract Documents – 3,585 pages".]
Joe Noss: The last piece I'd add here is that the goal is ultimately, once this data's all in place, for anyone to build whatever agents they want. An agent in this context just means automated workflow, right?
[00:19:43 All three options are selected and "Next" is clicked on, leading to a new page with the text:
"What should the agent do with this data?
Filter & Match – By vendor, department, value
Classify Ownership – Canadian vs. Foreign (UBO)
Trend Analysis – YOY growth, renewals
Rank & Compare – Top vendors, concentration".]
Joe Noss: Define the data inputs you want, define whatever are the types of enrichments or processes you're looking for.
[00:19:48 "Filter & Match", "Classify Ownership", and "Rank & Compare" are selected and "Next" is clicked on, leading to a new page with the text:
"How should results be delivered?
Executive Report – PDF with charts & tables
Email Digest – Summary to your team
Live Dashboard – Auto-updating view
Slack/Teams Alert – Real-time notifications".]
Joe Noss: The output, be it an executive report or an e-mail.
[00:19:53 "Executive Report" and "Email Digest" are selected and "Next" is clicked on, leading to a new page with the text: "Schedule: Daily, Weekly, Biweekly, Monthly"]
Joe Noss: And ultimately, the frequency you want and where you need that to be delivered to, and this is the sort of thing where the expectation is not for you to do this all yourself, right? We will help you and make sure every piece of the puzzle is aligned, but this is a demonstration of how you can start to use this data to actually replace some of the more tedious tasks you may have.
Taki Sarantakis: Now, Joe, obviously, with all of this capability, with all of this capacity to generate instant insight, clearly, you have, what, 30,000 employees, 40,000 employees, 50,000 employees?
Joe Noss: Yeah. So, we're at five folks right now. We're growing quickly, but I think it's a demonstration of the power of A.I. today where the thing it has the greatest impact on is our ability to code and create code and move very quickly in that context, and I think something that's challenging, and this is true for governments across the world, is IT spending for government is only going up while we see in the stock market that all software is being devalued because people know it's a lot cheaper to build, right? So, how do we help government realize these gains of coding in software being cheaper? And to me, I believe a critical part is just taking all these data inputs, cleaning them up, and getting them A.I.-ready so in two years, folks won't even have to work with me and they can just create these types of reports, workflows, dashboards, applications themselves. The first piece is just getting all this data clean and ready and ultimately free from the legacy ERP systems that trap them in, right?
Taki Sarantakis: I'm actually going to, if I can, if I can be so bold, I'm going to correct you. The first step is actually digitizing. You would be amazed how much data in governments, including the Government of Canada, has not yet been digitized. And at one point, I certainly hope it's less now but at one point, the footprint of the Government of Canada in terms of our holdings for offices was about 7% filing cabinets. So, if you think of that, if you have 100,000 square feet of office, 7,000 of that used to be filing cabinets. So, we have a lot of data and we generate data every single day.
Joe, thank you so much. This was a wonderful demonstration. What I really like is how quickly you get the insight, whether you're a manager, whether you're a director, whether you're a deputy minister, whether you're a minister, whether you are a parliamentarian. This is the type of information that should be accessible nearly instantly to decision-makers because it's factual information. It's not information that requires analysis. It's not information that requires coordination. It's simply facts. And for me, getting public policy right means first having your facts right. And until then, unless you have your facts right, you cannot move forward and do intelligent public policy. Thank you so much for kicking off our CSPS tech demo series and I look forward to watching great things from you and your other four people in the years to come.
Joe Noss: Yeah, no, thank you. I appreciate it.
Taki Sarantakis: Take good care.
[00:23:38 The CSPS logo appears on screen.]
[00:23:43 The Government of Canada logo appears on screen.]
