The following transcript was generated by AI and may contain inaccuracies.
Matt Goeckel: Today we’re talking about Genesis, our agentic AI solution that we recently released and recently made available for the private sector, which is where most of you on this call come from. If you’re doing internal investigations, e-discovery, fraud work, incident response, due diligence, or really any type of investigation, and you feel like you’re drowning in evidence, this session is built for you. Hopefully we’ll be able to give you an avenue to help deal with the mountain of evidence, particularly digital evidence, that you’re facing today.
My name is Matt Goeckel, Director of Technical Marketing here at Cellebrite. Prior to joining Cellebrite, I spent about 18 years in law enforcement, including roughly a decade leading a digital forensics lab down here in Florida, where I still live today. I don’t want to read from this, but apparently my firsthand experience gives me deep practitioner-level insight into how investigators can use tools like Genesis to work faster and more effectively.
All of which is to say: I’ve done the job. I did it for quite some time, and I’ve been here with Cellebrite for several years now. So I’ve seen how this industry has evolved, and how the amount of data we’re all dealing with has continued to grow. I know how hard it was to manage that data years ago, and today it’s gotten even harder.
Let’s jump into it. A few things we hope you walk away with today. First and foremost, the forces reshaping enterprise investigations: evidence volume, mobile data, and the fragmented sources of data we’re bringing in — logs, CCTV files, statements, reports, emails, whatever it is.
We have so many different sources of data, and so much of it, that review teams are getting to a breaking point. We’ll talk about how agentic AI can help with that, how it will accelerate evidence triage and evidence analysis in general, and how it will help you surface leads, connections and timelines.
We’ll do that through the slides, and then I’ve got a short video where I’ll run through the Genesis product so you can see what it looks like and how it operates. And then, of course, the thing I try to drive home every time I do one of these webinars or talk about Genesis is the human-in-the-loop aspect, which helps with the defensibility of an AI product.
We’ll see how source traceability and guardrails come into play there, and the governance we’ve put in place to help those of you who have to deal with legal and compliance — which is probably all of us — be more successful with your investigations.
With that, before we kick off, let’s do a poll. We’ve got a couple of these to keep everybody engaged as we go. The first one: what best describes your AI expertise? How do you feel about AI? Some of you are maybe still in that “AI is scary” camp. Hopefully that’s a minimal number of people here, but if you are, welcome.
Hopefully this will help demystify artificial intelligence a little. The one thing I can give you is that it’s not going anywhere. Most of you, I’d imagine, are going to fall into the intermediate bracket, and slightly fewer into the expert space — this is all me guessing — where we’re starting to feel more comfortable with AI, or maybe we’re building with it and creating apps and automations to make our lives a little easier.
And then someone here, and I know I always have at least one, is going to be an AI genius, running their own open-weight models on a Mac Studio or whatever computer they’ve bought, doing all the custom training and all of that fun stuff. With that said, let’s see what we’ve got.
I was close. I think I was spot on. There are a few of you still learning how to navigate AI. It is a scary concept, and the speed at which AI has taken over the world is certainly intimidating — a daunting thing to think about and look at, and the stuff it’s done.
Hopefully after this we will demystify it a little for you, and you’ll see how it really can be used to help you in your investigations and your day-to-day work, to do your job a little easier and a little faster. Then the good bulk of you feel like you’re intermediate, and we have a couple of experts. Like I said, there’s 1.6. I don’t know what the full number of respondents is on here, but that would be approximately one or two and a half of you who are AI geniuses. So welcome to you as well.
With that, let’s jump into an AI overview. What I hope to do here is aimed more at the new-to-AI and intermediate folks — again, to demystify a little, make you feel more comfortable about what AI is, and also to explain why some of the things that make AI scarier, like hallucinations, actually happen, and how we at Cellebrite are working to minimise them.
You’re not going to be able to go and take a Claude certification test when we’re done with this webinar, but it will give you a basic foundation in how AI operates.
First and foremost, the thing we want to point out is that AI is not one thing. It’s a broad term, and depending on who you talk to it could mean machine learning, automation, image recognition, translation, generative AI or agentic AI. That’s part of why the conversation gets confusing.
Even the formal definition we have on the screen on the right, which is from the EU AI Act, describes artificial intelligence very broadly. So today we’ll narrow the lens. We’ll focus on what I’d say are the newer capabilities getting the most attention right now, especially in investigations: generative and agentic AI.
These are the tools that can summarise and generate responses. They can help you ask questions across large data sets, surface possible leads, or even suggest next steps. And because of the usefulness of these tools, this is also what raises the same questions those of you in the “AI is scary” box are asking: How can we trust this? How can we verify it? How do we know it’s handling our data properly? We’ll get into that as we move along.
With that shared understanding, let’s talk about the history of AI and how we got to where we are. Artificial intelligence as a concept has been around since the 1950s, which lines up with when Alan Turing developed the Turing test — the standard for evaluating a machine’s ability to exhibit intelligent behaviour equivalent to a human. That’s where AI got its start. It’s been around a long time.
In the 1980s we saw the rise of machine learning: the art of training a computer to recognise, for example, images using data, something we can improve over time with more training and reinforcement.
By way of example, we’ve been using machine learning in our Pathfinder product since it was released in the mid-2010s. We use it primarily for image and video recognition. What we did was sit down and train a computer to understand what drugs are, or what a weapon is, or what an ID is. You continuously show the computer more and more of those images, and eventually it’s able to work it out on its own.
Let’s go to an easier example: a car. I know what a car looks like, and now I can show you pictures of cars. That iterative process makes it stronger and stronger. So something that’s complex and technical and feels really new age has now been around for 40-odd years. I was born in the ’80s, so I hate to think that’s a reality.
Then we get into the 2010s and deep learning, which is technically a subset of machine learning. This is the concept that powers the capabilities we’ll focus on today: generative and agentic AI. These are tools that summarise and generate responses, and this is the world we’re currently living in — the 2020s of generative and agentic AI.
These large language models — or sometimes, for vision work, you’ll hear about vision language models — have the ability to actually generate an output and give it back to you based on the training they’ve had.
And then, of course, who knows what’s going to happen when we reach artificial general intelligence. That’s the AI some of us are excitedly waiting for and some of us are convinced is going to be a recreation of the “Terminator” movies. It’s the AI that understands, learns and applies intellectual tasks across any domain. The thought with AGI is that it will at least match, if not surpass, human cognitive abilities. But that’s down the road — which, at the rate AI is developing, could be next week or could be the 2030s. We’ll see. For today we’ll concentrate on generative and agentic AI.
One thing to really think about, again for those of you newer to AI, is that it’s basically a sophisticated multi-step autocomplete. When we go through the process of training artificial intelligence, it’s learning off all of these things we’ve pulled from books and the internet — it could be anything from a post on a user forum to a book.
Eventually it’s learning what the most likely next word is. If you were to prompt the AI with “peanut butter and ___” and tell it to fill in the blank, the most likely response is going to be “jelly”. The AI is going to look at the next possible words — technically tokens — to follow “peanut butter”. Something like “bread” is going to score relatively high, because peanut butter and bread are associated. But “jelly” is going to be, “Hey, this is probably the most likely thing, so that’s the answer I’m going to give you.”
Same concept for “the sky is ___”. Blue is going to score very high. Green, not so much. Pizza is going to score at the bottom — “the sky is pizza” probably isn’t going to score very high. So the most likely response you’ll get is “the sky is blue”.
What this translates to for us is the magic of artificial intelligence: those predictions have become extraordinarily good. Any of you who have used a current model, especially the foundation or frontier models like Claude, OpenAI, Grok and Gemini, know it can be really good at what it does.
But the limit is that the AI is not storing facts. It’s not storing the fact that the sky is blue; it’s generating that from a learned pattern. What that leads to is AI’s ability to sound completely confident and yet be completely wrong in some circumstances. This is the concept of hallucinations that we hear about.
They can be reduced, but realistically they’re never going to be 100% eliminated. There are ways to mitigate those issues, and we do that at Cellebrite through our models — ensuring we’re using the best available models, and giving them the training we’re able to give them — and also through sophisticated system prompting, to make sure any hallucinating is minimised.
We’ll also see, as we get into Genesis, that one of our requirements is that we provide the sourcing back to you as the user. So in the rare event — and I say rare because I’ve been using Genesis now for almost a year — that it does hallucinate, you’re able to verify it through the data itself.
With those concepts in mind, let’s talk about large language models versus AI agents, because this is an important distinction for today’s session. With large language models, the things most of us are using every day — those of you who have a Claude window or whatever open on the next screen over, as I may or may not — I send a prompt, it processes that data, and it gives me a response. Summarise this document, translate this text, explain this topic, answer this question. One shot: question in, output out.
An AI agent works differently. It takes a goal — not just a question, but a goal — and runs it in a loop. It plans, it understands what the request is, and based on that it takes actions, using tools and skills to go and get that information. Once that information is returned, it adapts. It checks its work and iterates, to do its best job to make sure it’s correct. Or, if it’s a multi-step request, it takes all of those steps before it gives you the answer back.
That’s how we see agents researching topics across multiple sources, analysing large data sets — hint, like we’re talking about today — or completing complex multi-step tasks.
Genesis is built with that agentic AI in mind. It’s built as an agent, not as a chatbot. That’s the reason we’re able to work across an entire case: we pull from the UFDR, and then we pull from documents, call records and media all at once, instead of you asking one question about one file at a time and getting one response. That’s a big difference, and I want to make sure we’re clear on it.
To that end, the next poll: what is your organisation’s current AI stance or policy for investigative work? That could be anything from “what policy?”, as in we haven’t even thought about it yet, all the way up to “we’re working on it”, or “we have a policy but it limits what we’re able to do” — which I’d expect a lot of corporations to fall into. It’s, “Okay, you can use AI, but don’t go too ham with it.”
Then some of you, hopefully very few, may have a policy in place that says no AI is allowed. That still falls in line with the “AI is scary” mindset. Improperly implemented and improperly used, it certainly can be a risk factor. But when used and implemented properly, as you’ll see, AI will help you do your job quite a bit better. I’ll give a few more of you a quick second to answer the poll.
We’ve got a few results here. Again — gosh, Matt, it’s like I’ve done this poll before with other groups. Thirty per cent of you don’t have a policy in place, which is neither here nor there, but at some point you’re going to want one, depending on the size of your organisation. If you’re a one-man shop and your brain is your AI policy, fine.
But if you’re a larger organisation, it’s probably a good idea to have something in place, just to keep people from throwing things like PII up into Claude or ChatGPT and letting the next generation of AI train on your data.
Four per cent of you are not able to use AI. Hopefully something like this — and you’re welcome to take this recording and show it to your boss — or some resource somewhere can help management or IT, whoever made that decision, change their mind and take another look at the usefulness of AI when applied properly.
Most of you, as expected, do have a policy in place and are allowed to use AI to a degree. That tracks with what we see across all industries, and I think it’s the right place to be: use AI, be smart about it, be intentional, decide what you put into it and where.
For example, with Genesis I’ll put all of my case data in there, but I’m not going to have it generate a picture. It actually won’t do it — it’s prohibited from doing that. That’s not what that model is built for, and that’s not what that data set is built for.
On the contrary, I also would not throw my case data into Claude and then ask it questions about that case. Number one, it doesn’t necessarily understand DFIR. But number two, even though I have training turned off, who’s to say whether that data is making its way back to Anthropic or not? I don’t know. So I’m going to be safe rather than sorry.
With that, let’s move along to the reality of enterprise investigations. Most of you are aware that we do a couple of industry trend surveys every year — one for the private sector, one for the public sector — where we ask about what you’re all seeing and experiencing in your industry. Let’s take a look at some of those results.
66% of enterprise cases now involve a mobile device. That’s a six-point increase from 2024. So we know mobile data as a whole is becoming more and more common in enterprise cases, especially for those of you in larger corporations with managed devices or bring-your-own-device environments. There’s always a chance a mobile device is going to be involved in your investigation.
54% of respondents say e-discovery is the single most important investigative use case, and 54% cite collecting evidence from chat and messaging apps as their top investigative challenge. That’s true across the board with chat and messaging apps.
Speaking from experience, and from conversations I have with state and local agencies, e-discovery isn’t really a problem for them — but those other challenges really resonate across the board. It’s always interesting to me to see how the different industries are so similar and yet can be so different.
Some of the issues you talked about in that survey, and other things we hear, bring us to this thesis: the problem isn’t that there’s a lack of data. I don’t think any of us on this call would disagree and say, “Oh no, I don’t have enough data.” I think we all have enough data.
The phone sitting in front of me is a two-terabyte device — more, incidentally, than the computer I’m talking to you on right now. If you’d said that was going to be a reality even 10 or 15 years ago, I’d have said you were crazy. But here we are.
So we know we don’t have an issue with data. What that leads to are manual review cycles that slow investigations down and delay critical decisions. As you try to go through all of this data by hand, how long is that going to take, and what’s the risk that you miss something?
Case backlogs are growing, and it all relates back to trying to look at this stuff manually. There’s so much of it that the next case has to sit and wait for somebody to become available to extract the data, if it’s a digital device, then process it, then go through it and dig through it and understand databases and all of that.
It winds up with key leads being buried. We have fragmented evidence, we have multi-source evidence, we’re trying to go through it all by hand, we’re copying things out to Excel spreadsheets and building tables and different views to make sense of it — and we end up missing the one key thing that would either help us close the matter or push it in the right direction.
What that leads to is relying on more headcount, or even having to hire outside firms just to keep up with the amount of data coming in. It’s a domino effect. It just worsens and worsens, and it ends up increasing your risk.
To look at this as a chain: we talk about limited resources, and when you have limited resources and you’re trying to do it by hand, you end up in a time-consuming evidence review loop. You’re trying to go through the data of one case, so your production isn’t as high as it could be. You’re not finding the things you need to find, so you’re losing insights, and that leads to a risk of compounding exposure.
Every step in the chain makes the one before it a little bit worse. It compounds. That’s the cost of not solving the problem. It’s not shocking or new to any of us — we’ve been experiencing these issues since the beginning of time. But it is getting worse, because number one, we get more and more sources of data, and number two, those same sources get bigger and bigger. The problem will continue to compound as that trend continues, and we all know it will. We all miss the days of eight-gigabyte iPhones, which we thought were big then, and now we have terabytes of data.
I want to talk very briefly about something I think is really important. It matters to me personally, because I ran a lab for a long time and did this work, but it also matters to me as a human being. One of Cellebrite’s baked-in principles around AI is that we view AI as an assistant, not a replacement for investigative judgement.
The idea is that AI helps you do your job a little more easily. It accelerates triage, it accelerates finding key evidence and leads, and it helps you connect evidence across devices and formats. We actually had a question about using Genesis to look at different users and collate that data together — you can absolutely do that, and it’s very helpful. It also builds case timelines automatically.
But what we leave with you as the user is the interpretation of what that evidence actually means. Genesis will give you an output — we’ll see an example — but it’s still going to give you the sourcing, and it’s going to leave it to you to say, “Okay, yes, this is correct. The AI reported it correctly,” and to decide what goes into the case record and what’s a valid lead. You get to make that final call.
That little subpoint at the bottom: Genesis accelerates the work, your investigators still make the call. I point that out because I think it’s important to hit home, and also to drive home as a concept from Cellebrite as a company. I don’t have all 10 principles memorised, but human in the loop is, I believe, number one on that list. It’s very important to us.
And why not use general-purpose AI models for this? I touched on it earlier when I was talking about dumping my cases into Claude and why that’s a bad idea. General-purpose models don’t have investigative context. They don’t have forensic workflows. They generally don’t provide traceability — I’ve had to do extra prompting to get sourcing back on documents I’ve had them do outside research on. And they require a lot of additional controls, especially if you’re on a personal AI plan, where there’s no data use agreement in place, realistically.
You can turn off training, but all of that data is still being passed into the same Anthropic that my personal Claude account is passed through, and everybody else’s. So there’s no real security or safety around that. And there really is a difference between working digital evidence and understanding it.
Responsible AI for us in this investigative space looks like source-backed findings. It’s keeping investigator oversight, keeping security and privacy controls in place, clear operational boundaries and transparent outputs. We’ll see the example when we run through the video: I’ve asked Genesis a question, it’s given me an answer, but it’s also given you the source it got that answer from.
It does that every time, because we want you to be able to look at it — whether you’re the investigator who needs to give it back to your extraction team or your digital forensics expert and say, “Hey, I found evidence that this person was stealing our IP and giving it to the competition. Can you verify this against the raw extraction?”
Or you’re the examiner yourself. We’ve had examiners who have used this come back and say, “All day long, I would rather let Genesis go out and find these potential items of evidence, and then I just go back and verify it, rather than having to go through and try to find this stuff manually myself.” It’s that time-saving factor.
Before we jump into the Genesis demo, one more quick poll: your team and AI in investigations. Are you using it as an approved, everyday part of your workflow? Do you use it in limited ways? Are you thinking about it? Matt’s going to venture a guess again, and I might be wrong on this one, but I’d guess you may be exploring or piloting AI tools, and that’s why you joined this webinar. Or some of you aren’t doing it at all yet, which is also fine, and also why you’re here — so you can get that information. I’ll give you a second to answer.
Let’s see what we’ve got. Wow, very even across the board. Most of you are using AI in limited or approved ways, and then a good quarter of you are exploring or piloting AI tools. Another roughly quarter aren’t quite there yet. For that 50%, this is hopefully the wake-up call to help you understand — or help your management understand — how this can be beneficial.
For those of you using AI every day, or even in limited ways, a lot of this is probably a bit of a review. “Okay, yeah, Matt, I know we’re doing that. Now show me the goods” — which I’m about to do for you in just a minute. I’ve got a couple of slides before that, sorry. But hopefully you’ll see how something like Genesis can fit into your workflow and help you do your job a little more easily.
Let’s talk about the product. Let’s talk about why we’re here today. Genesis turns complex digital evidence into intelligence, fast. It’s an agentic AI system built specifically for investigations. It’s built to understand investigations, and that’s really all it does. Like I said, you can’t go into Genesis and say, “Hey, make me a picture of a bag of popcorn.” It’s just going to say, “No, sorry, can’t do that. It’s not my job.” It’s just for you.
It supports a multitude of case file formats — over 35. From a forensic standpoint, UFDRs and portable cases. For those of you on the call saying, “Hey, what about computers?”, that is on the roadmap and we’re developing it now, so we should see it released relatively soon.
Call detail records — it does a really good job with things like that — and structured logs and returns. From there it’s your common document, image, audio and video formats that we’d typically come across. Genesis can process all of those. It will read and transcribe documents. It will listen to and transcribe video and audio.
And if you have a thousand files for a case, just throw them in a zip file and give it to Genesis, and Genesis will open it up and process it for you.
Like I said, it’s built for investigative work. We have AI-powered evidence analysis and cross-source correlation. As I mentioned, one of the questions right off the bat — and I love that the questions are coming in — was about that. We do look at the case from a holistic standpoint. We zoom out, and you can tell Genesis what part of the case you want it to look at, but whatever you provide and include in that prompt, it will look at for you.
I’ve hit on source traceability a couple of times; we’ll see it live. Investigation-aware workflows: the whole concept of Genesis was designed around conducting an investigation. And then secure, governed deployment — I think we have another slide on that in a minute. Everything is fully isolated. Your data is your data. We don’t cross paths, Cellebrite employees cannot access it, and we never use your data to train our AI models. It’s a non-starter for us, and hopefully that gives everybody more of a comfort level with using AI.
The workflow: prioritise what matters. Think about what you need to know. It’s just like prompting a normal LLM. Okay, I want to know something about this investigation — take it out of your brain, type it into the computer, push the enter button, and let Genesis go off and cook for a little bit, and it’ll bring back those answers. It will help you connect evidence across your case, and validation will be front and centre.
With that, let’s see Genesis in action. This comes up in the video window, so you may have to open that window if you don’t see anything. I’ll be back with you in a few minutes.
Matt Goeckel: Here we are in the Genesis interface. If any of you have ever used an AI chatbot before, this will look familiar: it’s a screen with a chat box, which is most of everything you need to know. Genesis is a little different in that it’s not for general chat — it’s specifically for querying evidence or data in your case.
There are two steps that are a bit different from what we normally do. Number one, we’ll select a case. In this case I’m going to select this Capture the Flag from 2025. Number two is to pick which artifacts you want to query against, if you need to change them. Here we have three UFDRs. If I wanted to only look at two UFDRs, or if I have a bunch of documents I want to look at, I can select that there, and that will limit what Genesis is actually looking at. From there we can start prompting and querying the evidence.
Let’s look at an example. If we say something like, “Is there any evidence of IP theft or data exfiltration in this case?” — if I could spell correctly — it’s that simple. It’s your brain to the computer. Whatever you’re looking into, whatever information you want, you put it into the prompt, hit enter, and give it a few minutes to get the answer.
While Genesis spins up all of its agents and starts doing some work, I’m going to cover a couple of other things. I’ll go back to a new chat and jump into this case here. What I’d like to point out is this capability called Deep Investigator.
This is really handy. Any of you who are more forensically inclined, or who need to go deeper into the data, can turn on Deep Investigator and then start asking Genesis to process additional databases. These may be outside the normal decoded databases you’d see, but if you need to get a bit deeper, this is how you can do it.
As you can see, we have all the common databases we’d typically dive into in an investigation. We can also do a browse and ingest. If I was interested in, let’s say, consolidated, I could go here — it’s under the caches folder — and if I click on it I can see consolidated.db is available.
The other nice thing we’ll do for you is tell you exactly what that database does. Then I can click it, click ingest, and it will bring in this database, and now I can query against it as well. That’s really useful, because I don’t have to understand the database off the bat. Genesis can do that for me.
Then I get the results, and in a second we’ll look at those and see how we can take that returned data, export it, and give it back to the examiner in the lab for the verification and validation process.
With that said, I see our green notification down here saying our prompt is done. We’ll hit view, get out of Deep Investigator, and we can see that yes, in fact, there is substantial evidence of covert file transfers, data exfiltration and acquisition of sensitive proprietary leverage assets across the device extractions.
What Genesis will do at this point is tell us what it’s found that it thinks is relevant to the prompt. We can see a spreadsheet transfer, and we can see that a WhatsApp media file containing a data spreadsheet with numerical rows was recovered from Kevin Malroy’s device, and so on.
Not only is this giving us a plain-language view into not just one UFDR but, in this case, three UFDRs — and if this were three UFDRs plus twenty-five documents and some video and a few other things related to the matter we’re working, everything would be here in one place.
What you can see is that each of these little icons is Genesis telling us where it’s sourced information, where it’s getting the information it’s presenting. When it says, “I found a message that said, ‘Sent you a spreadsheet, oops, wrong person'”, we click this and it brings up a side panel that shows us the chat where it came from.
If we work our way down here — there it is. “Sent you a spreadsheet, oops, wrong person.” So we’re getting the information. We can go to the top and see exactly where it was pulled from. We can see who was in the chat. We even get the chat reference ID.
Secondary to that: once I find information I’m interested in — let’s say this is important to our matter and we need to dig into it a bit more — I can tag it, just as I would in Physical Analyzer or Reader. Tag: we need to follow up on this, whatever we want to say, hit Confirm, and now this item is tagged. It’ll live over here in our tagged database, which I’ll look at in a second.
We can go down here — yes, we have an image, and here’s our media summary: a close-up view of a piece of lined notebook paper with handwritten text in black ink.
The important thing to point out is that the sourcing is paramount. It’s part of that human in the loop I’ve been talking about. But also, for images, videos and audio files, Genesis will analyse the content of that media and give you back a summary. If it were a video, it would give us the video with the transcript. If it’s audio, we get the transcript. With an image, it gives us a description of what it sees.
Additionally, it’s going to show us artifact details. Up here it’s explaining what it’s giving to us; down here it’s giving us an artifact-by-artifact explanation. We get a confidence assessment of the quality of the evidence it found, and then a potential next recommended action.
This is optional — you can follow it, you don’t have to — but if you’re stuck, it might be a good way to bump the investigation forward. It’s context-dependent, so whatever we’ve been talking to Genesis about, it will return a next recommended action based on that prompt.
The other thing we can do is use one of our widgets to create custom views. But I want to hit a couple of things first. Number one is the ability to export. I could export this prompt as a PDF. I can also export the entirety of the session — the entire chat — and I can share it.
If I decide I want to share this with, let’s say, Evyatar, I hit Share, and Evi will get notified and be able to access it. I can also give feedback: this was good, this was bad. Or if I want to give more feedback, I have that over here.
Let’s kick off this timeline view, which gives us a chance to see the visual artifacts Genesis is capable of producing. Now I’m asking — again, prompt-specific, based on my prior prompt — “Create a timeline of data exfiltration activities between Kevin and Ann.” That’s going to cook.
The other thing I want to hit is the ability to tag evidence. I showed you a second ago how I can tag evidence. Now that I have that, I can hit Export Case Tags. That’s going to generate what’s called a .gen file. We’ll hit Save, and that saves out. I can take this into Physical Analyzer or Reader and load it in, very similar to a session reviewer file, and we can see the tags I’ve made.
This is a small item — 2.8 kilobytes — that I can attach to an email and say, “Hey, please review these tags. This is the stuff that’s important to my matter. Can you look into this and verify and validate the information?”
While I was talking about that, our timeline finished, and if we scroll down we get a nice visual timeline of events. We can pop it open and make it a bit bigger. It even found a spreadsheet in a WhatsApp media file that was shared, which is maybe related to this case, maybe not. We’d want to dig into that. But of course, if it’s something we’d never seen before, this could be a new lead in our case, or perhaps even the evidence we’re looking for to resolve the matter.
That basically sums up the system. It’s very powerful, but it’s also very easy to use. The last thing I’d point out is that among all of these cases, I can share case access if I need to. Let’s say this is a Colin DeCops 2024 case that I’ve uploaded and I want Andy to have access to it. I’ll select Andy — Andy has two users in here, so we’ll select both — and hit Share, and now Andy will get access. I’ll go back to the chat, refresh, back to the home page, and we’ll go back to the webinar.
Matt Goeckel: I’m back. Very quickly — I’ve just realised we’re right about out of time, so I appreciate you sticking with me. I’m almost done, I promise. Thankfully I’m just so entertaining that you all want to stick around forever and ever.
A couple of things I wanted to double-highlight. Per-tenant data isolation — and I saw the question come in, Genesis is SaaS, it’s cloud-hosted, so as far as deployment goes it’s all fairly easy. Everything is logged, so we have full audit logging, and at the same time we will never train on your data.
The last thing I want to skip ahead to is a customer success story from one of our early access private sector users over at Interpath. I’ll leave the slide with you — if you want to read the whole thing you can — but I just want to read that last quote: “We worked a case related to potential financial wrongdoing in an IPO that was ultimately scrapped, and it took us months to manually piece together who was involved. Yet when we ran all the data through Genesis, it took minutes to surface the same information. The speed and accuracy amazed us and helped validate our original findings.”
That’s the big takeaway I hope you take from today. Using a solution like Genesis to work through all of that data is going to help you find evidence that would otherwise take a couple of months of scrutiny, within a matter of minutes. That’s one success, and we’ve got a ton of success stories from our early users going through all of this data.
With that, my friends — again, I know we’re a couple of minutes over — thank you for sticking with me through this overview of Genesis and artificial intelligence here at Cellebrite. Don’t forget to sign up for the free trial. You have to use your work email, so Gmail is not going to make it through. But sign up, thank you for your time, and we will see you on the next one.
