CTO at Orbital Witness (AI for LegalTech). Built Appear Here (PropTech) and the Caffè Nero app (Fintech), advises startups and VCs, founded an AI company. Ships early and often. 🚲☕️🔧⏱🚀
Back in June 2023 I gave a presentation titled "Generative AI: Opportunities and risks for property transactions" in which I demoed an internal AI Assistant project that my team and I had been building ever since ChatGPT launched to the world in November 2022 and it became abundantly obvious this new Generative AI technology had the potential to transform the legal industry. We received a tremendous amount of engagement off the back of my presentation and much of it was from customers wanting to get their hands on this AI Assistant that we built so they could try it out for themselves. Because our initial AI Assistant was only an internal tool, we need to pull it out of our backend application and productionise it for law firms and other companies so they could begin trialing it and giving us feedback.
Fast forward a few months once we had ironed out a number of challenging technical problems with building a bleeding edge AI system built on top of the latest large language models (LLMs), our own customised prompt engineering, a retrieval-augmented generation (RAG) system and our own proprietary legal specific document OCR and structuring technology all the while controlling for the eye-watering costs of some of the more powerful LLMs like OpenAI's GPT-4.
Today BCLP and Orbital Witness have publicly released the details of our "global design partnership to test, optimize and deploy Orbital Witness’ lease reporting capabilities in its latest generative artificial intelligence (AI) solution, Orbital Copilot, across BCLP’s U.K. and U.S. real estate practices." I'm proud to say that the AI Assistant I showcased at my presentation back in June 2023 has now taken on a life of it's own as Orbital Copilot and has a ton more functionality, improved accuracy, ability to work with any legal documents and is initially focused on lease reporting for lawyers and legal professionals as we build this out for BCLP and others. The press release goes on to say: "Orbital Copilot is the first real estate-sector-specific AI assistant created for real estate lawyers that is capable of analyzing and reporting on any real estate document. By leveraging Orbital Witness’ market-leading capabilities in real estate-specific AI and the deep sector knowledge and experience of BCLP’s global real estate practice, this first-in-market collaboration seeks to use this cutting-edge technology to enhance BLCP’s reporting on complex commercial leases, providing accelerated insights to their clients."
I'm also quoted in the press release alongside Samant Narula, BCLP’s U.K. head of real estate:
Andrew Thompson, chief technology officer at Orbital Witness, says, “A key part of our strategy with Orbital Copilot is to develop real estate-domain-specific AI that enables us to deliver solutions for property professionals that are incomparable in terms of capability, quality and reliability. This collaboration and pilot program with BCLP allows us to demonstrate and test that value at a firm that shares our ambitions and vision for this technology.”
BCLP has one of the largest legal real estate practises in the country, so they're a brilliant law firm to partner with for the ground breaking Generative AI product called Orbital Copilot that we've just built and launched to the world.
On June 14, 2023, Orbital Witness held an event for #LondonTechWeek to explore how Generative AI will revolutionise property due diligence. I gave the keynote presentation to a room of partners and innovation specialists from commercial law firms, residential conveyancing firms and LegalTech companies.
Here is the full 38min video of that presentation along a product demo at the 32min mark. The product I demoed is an internal tool my team of data scientists and legal engineers built using a GPT-4 based agent and vector search on top of legal documents from HMLR such as title registers and scanned PDFs of leases and deeds of variation:
Here is the complete slide deck I presented for my keynote with all transitions intact (PDF version at the end of this blog post):
Here are all the references associated with various slides:
Given my current ongoing research on the topic, I'm going to use this blog post as a curated collection of the best content I've read across a variety of themes related to LLMs / ChatGPT / Dall-E 2 / etc. I'll update it regularly with links to new blog posts, podcasts, videos and anything else I find particularly compelling. The date in parentheses is when I added the link to this blog post and over time it will provide a change-log for easier reading.
"Larger context means more complex applications: A 32k tokens context means about 25 000 words or 50 pages! This is huge! It means it can process full reports, short books or long chapters, taxation code, and so on, enabling scenarios that were impossible with ChatGPT’s 4k token limitation!"
Ethan's Twitter (he's a Professor at Wharton studying innovation & startups) has been one of the best resources to see what kinds of things are possible with ChatGPT
What does 'flexible working' actually look like at our company?
This is a repost from Orbital Witness's Tech Blog that I recently setup with the engineering and data science teams to both recognise the work colleagues were doing internally as well as promote their work externally to attract potential candidates when hiring.
Clarifying the ambiguity
Remote working, working from home (WFH), hybrid working, flexible working, cross-border working, mandatory days, offsites, in-person, distributed… on and on the list goes 🤔. There’s a plethora of ways companies and employees are defining how they want to work and what’s expected of everyone while at work. Depending on your previous work environment or what you’ve read about how other people are “working flexibly”, it can mean many different things to different people at different times.
Having a blanket policy that works in all situations for all people is difficult to say the least. Instead of having such a policy that covers all possible scenarios, we instead took the approach to gather feedback from employees and combine that with what we value and want to foster as a company long-term. From this more organic approach, a set of day-to-day flexible working practices have emerged which employees find works well for them both personally and professionally. This blog post aims to highlight many of those flexible working practices; both to codify them for our own internal use but also to share them with potential future employees who want a sneak peek into how we work.
Asking employees for their feedback
In May 2021, while the UK was still under a partial lockdown, we surveyed employees about what they’d like to see in the next 6 months in terms of flexible working. We wanted to use the responses to create a consistent company policy that worked for everyone. We wanted to learn more about what we had been hearing from some employees: that they were missing elements of a working environment when forced to work from home for long periods of time during the pandemic. Here are the survey results off the back of asking everyone in the company this question: Over the first 6 months from lockdown lifting (21 June, 2021), how many days a week, on average, would you ideally want to be in the office?:
Along with collecting the quantitative data above 📊, we asked a few follow-up questions about what people value most about working from home and working in the office to help us understand how we could foster the right working environment for both settings. We created the word clouds seen below from everyone’s responses which showed us the different aspects of each environment that are valued by our current employees (the bigger the word, the more often it was mentioned by everyone):
Reasons for working from home
Reasons for working in the office
What we value and what we’re trying to foster
At the heart of our culture and values - most notably, Power to Our People - is the trust we place in our employees. We enjoy giving everyone both the autonomy to decide which work environment is the most conducive for them to do their best work, and the flexibility 🧘♀️🙆♂️ to put this into practice today, tomorrow or in the future.
At the same time, we also want to foster a thriving, collaborative, and pleasant working environment for those employees that have a specific need or desire for home or office working. Each team and each person works differently depending on their needs. Sometimes, the number of days worked from home or in the office may be a consistent trend, but it may also become sporadic and less predictable due to personal circumstances or the nature of the specific project they’re working on. The guiding principle we follow when allowing for flexibility is Customers => Colleagues => Self. This means that if we are ever in doubt about what to prioritise then whatever is best for the customer comes first, followed by what’s best for our colleagues as a team, followed by what’s best according to one’s personal preferences. This principle irons out most issues and allows everyone to work as flexibly as they need according to their task(s) while aligning everyone to what’s most important for the business to continue thriving.
The desire for flexibility was not only evident in much of the survey data shown previously, but also in one-on-one meetings with colleagues and in interviews where we listened 👂 to what potential candidates wanted in their future work setup and/or thought were missing from their current work environment. Pretty much everyone loves the comfort, convenience and solitude that home working provides but they also miss the structure/routine, contact with colleagues, and dynamism that multiple colleagues in an office can provide. At Orbital Witness we wanted to have our cake and eat it too 🍰 and therefore we actively foster the ability for any employee to choose either working option - on any given day. Sometimes people choose both options on the same day such as working from home in the morning and from the office in the afternoon!
Everyone is different and therefore requires a different setup to be productive and happy at their job. A 25 year old living in a cramped London flatshare 🏢 may need the respite and space provided by a quiet office environment with meeting rooms. However, a parent of two little ones 👶🧒 needing to conduct the school drop-off may find working from home or simply some more flexible time in the mornings really helpful to start the day off on the right foot. Then there’s also those times where someone just needs to find the time and space to get into the zone and spend hours of uninterrupted time solving a hard technical problem.
An added little bonus of working at Orbital Witness is that we currently have an office in a building operated by The Office Group (TOG) which is similar to a WeWork concept. Because of this affiliation with TOG, every Orbital Witness employee not only has their own home and our usual permanent office environment in Clerkenwell to work from but also 40+ TOG buildings throughout London (and even more offices in other cities in the UK and Europe) which they have access to with their TOG key card.
What’s actually happening day-to-day
Flexible days of the week
There isn’t a set number of days we ask everyone to be in the office and instead allow for more of an organic approach. So what’s actually going on in a given week? Here is data taken from our Officely Slack App which collects self-reported data from employees about when they’re coming into the office. Colleagues use Officely each week to inform everyone else about when they might be coming into the office on a specific day. Officely helps us all to more effectively coordinate meetings and have impromptu meetups between colleagues in the same team or across teams. Officely has been really helpful to make sure more of our one-on-one meetings happen in person over a coffee ☕️, while our communal lunches spill over many tables and after-work pub get-togethers with colleagues are a staple. Clearly, Wednesdays and Thursdays are good days to coordinate those things, with Friday being a little less busy:
Flexible location
The vast majority of employees at Orbital Witness are based in and around London with a few people as far flung as the seaside 🌊 in Kent and the countryside 🏕 near Bristol and Cardiff 🏴. As such, most employees come into the office on a weekly/fortnightly basis for a few or more days. Those who are more far flung come into the office monthly/bi-monthly for a few days at a time, often for our company all-hands once a month. We even have some colleagues who come into the office most of the week because they like the routine and structure of it. To summarise, there’s a complete mix of what people currently do and hence this is why we value continuing to have that flexibility. Our feedback and surveys show that it’s working well both in terms of productivity and employee happiness; therefore, we’d like that to continue 😊.
Given that we’re trying to foster both flexibility and an in-office culture, we’ve deliberately decided not to seek out fully remote colleagues who are not able to come into the office at all. We’re not dogmatic about this and will make the odd exception to the rule, but in the long-term we want to have a company with a healthy segment of employees who work and socialise in-person on a regular basis, especially while our teams grow and our culture evolves.
Flexible working hours
Our core working hours are 9:00am to 5:30pm but we’re not overly pernickety about following this and we certainly don’t monitor it. Many of the teams have stand-ups that begin at 9:30am and so long as people communicate effectively about when they might not be available during core hours, everyone is free to hop off to a doctor’s appointment or gym session throughout the day. Again, the reason for this is to give everyone flexibility; we fully trust our colleagues to balance both their personal and professional lives and everyone does just that. The reason we’re comfortable with this approach is that, from a management perspective, we follow what Andy Grove famously wrote about in his book, High Output Management, where he said measure output not activity. At Orbital Witness, we are also far more concerned with what output is being generated by a team/individual rather than tracking the activity or amount of time that produced it. As a result, we care less about the hours worked and more about the value we’re continuously shipping to our customers.
Company get-togethers
Once a month, we have our company All Hands meeting where we go over key performance metrics of the company, discuss one important topic that matters most and often have a fireside chat with a customer, investor, or industry leader. We’re always set up virtually for these meetings so that those who are working from home can participate. However, for the day of the All Hands, we encourage everyone across the company to be in the office where we have lunch together, schedule collaborative meetings within teams and have the company-wide All Hands meeting. Afterwards, we often have a planned social in the evening which has ranged from canoeing 🛶 Hackney Wick canal (and sometimes falling in) to competing in epic two-on-two ping pong 🏓 battles at Bounce to baking and decorating a Great British Bake Off style cake 🍰 in teams of two (and then devouring the cake before we’ve even got home for the night).
Other small niceties
Dogs in the office
Many colleagues decided to get a new family member during the pandemic 🐕 🐩 🦮 and we applauded them for it. We know how important dogs can be for many people and how much love and attention they need during the day. Because of this, people have the option to bring their dog into the office on select days so long as their furry friend is well trained and doesn’t disturb others who are hard at work. We created a little system so dog owners and some colleagues (who for various reasons may not want to be around dogs in the office) can easily coordinate which day(s) dogs may be in the office and everyone can work around this with little hassle.
Video-first meetings
The entire company has weekly stand-ups (Monday mornings) and stand-downs (Friday afternoons) where we briefly talk about the week ahead and/or what was accomplished, the tracking of our OKRs, and any big events/happenings that week. It’s a good touchstone for everyone to see the entire company each week and get a sense for what’s going on. Each team usually has a separate stand-up each day (focused on the specific details of their work) as well as other ad hoc product, design, engineering and/or one-on-one meetings. All of our meetings are set up so that colleagues can either video conference in 🧑💻 or be in the office around a single screen 📺 with good audio quality. As such, most people keep their cameras on as it’s nice to see everyone’s face during these meetings (unless of course people happen to be eating 🍔 or are having a bad hair day 👱♀️).
Working remotely for a period of time
We have a very international team and that means people’s families and loved ones are spread far and wide across the globe. Coming out of the pandemic, when travel was allowed again, it was vital that everyone was given the opportunity to see their family if they wanted or needed to. We had several colleagues fly off to their home countries such as Croatia 🇭🇷, Brazil 🇧🇷, Malaysia 🇲🇾 and Hungary 🇭🇺 for extended periods of time where they both took holidays and worked remotely for a number of weeks. This allowed them to spend much needed time with family and friends who they hadn’t seen for some time, while also enabling them to continue working on their projects without using up all their holiday allowance in one single shot. We have to be a bit careful to make sure everyone abides by UK tax laws (which dictate how much time someone can work outside the UK while being on a UK payroll), but so long as that’s satisfied, employees wanting to work abroad for up to 8 weeks a year is something that we support. We just ask that employees continue to overlap with the core working hours of their team. If you’re from France that’s a pretty easy trade-off but it can be more difficult the farther afield you travel (depending of course on your tolerance for changing when your work day starts and ends). Having said that, even one of our tech leads, who recently went to visit family and friends in Malaysia 🇲🇾, and one of our senior engineers, who recently went to Brazil 🇧🇷, didn’t mind too much about starting work in the late afternoon or early morning (respectively) to accommodate this.
The future is adaptable
Each team that works closely together is frequently fine-tuning their meetings and work practices to make sure they’re a good fit for everyone in the team. A team environment needs structure but also adaptability and we try to promote both. Our style of working, as mentioned above, highlights our core values around flexibility. In the future, things may change and our working practices may require further adjusting both as employee headcount grows but also as people’s expectations and wants/needs change. As a bold, ambitious and inclusive company, Orbital Witness aims to keep ahead of this curve by continuously speaking with current and future employees and adapting how we work best, while always putting first how we can maximise the value we deliver to our customers.
Codifying real estate legal expertise using data science and software engineering
This is a repost from Orbital Witness's Tech Blogthat I recently setup with the engineering and data science teams to both recognise the work colleagues were doing internally as well as promote their work externally to attract potential candidates when hiring. At Orbital Witness we're building Natural Language Processing (NLP) systems for the real estate legal industry.
Presentation Video (9min)
On October 5th, 2021 Andrew Thompson (CTO) and Brian Kennedy (Head of Legal Engineering & Innovation) gave a presentation at the LegalGeek conference in Brick Lane, London. The title of their presentation was Building software that ‘thinks’ like a lawyer. It was a short 9 minute presentation where we highlighted five hurdles that arise when building such a system.
Lawyers, around the UK and abroad, currently conduct due diligence for their clients every day so that land and property can be bought, sold and developed. One of the main limiting aspects of this diligence process is the time taken to read through all legal documents to identify potential risks that could delay or halt a transaction. At Orbital Witness we’re solving this challenging problem by combining real estate legal knowledge, data science (specifically OCR - Optical Character Recognition and NLP - Natural Language Processing) and software engineering to automate real estate due diligence.
Document OCR
To pick up on just one hurdle mentioned in the above presentation, Hurdle #1 was concerned with how unpredictable OCR can be on legal real estate documents that Orbital Witness consumes. As can be seen in the image below, leases are available in varying degrees of quality and can often contain a variety of interesting artefacts such as:
Both handwritten and typed textual content
Stamps that obscure textual content
Redacted sections of textual content
Struck out sections of textual content
Solving for these types of artefacts is one challenging step in a larger data pipeline. That pipeline starts with transporting and processing a collection of photocopied real estate documents and ends with real estate risks that have been accurately identified and classified according to the context of a transaction. This is no small feat to solve, from a data science and software engineering perspective, but we’re building a very capable team of lawyers, product managers, data scientists and software engineers to own this and other challenges and deliver solutions for them.
We’re hiring
If you are interested in the above challenges and are curious to know more about how we currently solve them, please see our open roles and get in touch with us via our Careers Page. If nothing quite matches your experience then still feel free to connect and message me, Andrew Thompson, directly on LinkedIn and I’d be happy to have a casual chat via video conference or over coffee ☕️.
Objectives and Key Results (or more commonly knowns as OKRs) are a well-defined way to codify a company's strategy along with tracking progress towards the intended results. John Doerr introduced OKRs at Google to great effect and has since written an insightful book called Measure What Matters that goes into the OKR framework and how it's successfully being used at various companies.
The OKR process can become overly complex rife with over analysing the data behind the numbers, poorly quantifying the key results and/or poorly communicating the ongoing impact they're having. I have seen many companies abandon the initiative part way through a quarter or even after a few quarters of failed attempts.
I've distilled the OKR process for product engineering teams and the wider company down to a simple tracker that can be used on a weekly basis. Below I will explain the "team tracker" along with some examples, followed by how each team's OKRs come together into a simple view across the company that anyone can understand.
Team OKR Tracker
The team OKR tracker is the starting point for the manager of each team. OKRs don't have a mandated length of time but I find doing them on a quarterly cadence allows enough time to make the upfront planning worth the time invested while still allowing for regular points throughout the year where strategic initiatives can be adjusted. Here is what a fresh version of the template looks like prior to a team's manager populating it with objectives, key results, owners and their associated weekly progress tracking:
There are several parts of the above sheet template which are worth digging into further:
Objective: This is what you want to accomplish. An objective should be aspirational, significant to the company, personally meaningful and ambitious. I'd recommend having no more than 5 but ideally 3 or less objectives per team (the old adage of less is more can't be said enough when setting objectives).
Key Result: This is how you will track the achievement of the objective. Key results should be easily measurable and limited in number. This is how the company measures progress towards its objectives while remaining flexible on the exact tactics to implement.
Owner: There should be a single owner for any given key result who is responsible for driving its success. In a product team this is often the CTO, Head of Product/Engineering/Design, Product Owner, Scrum Master, Tech Lead, etc.
Weekly Percentage Tracking: Each week throughout the quarter the owners of the OKRs should get together to discuss the progress of each one based on the percentage in the spreadsheet. Similar to a daily standup within a product team, any blockers or impediments should be raised and decisions made to resolve them.
Final: This is simply the final result of each OKR at the end of the quarter. Note that sometimes it may take a week or two in order to determine the steady state of actual metrics associated with a given OKR (e.g. waiting for A/B test results for a feature deployed on the last day of the quarter).
Forecast: This should be regularly discussed and re-forecasted if necessary to get an accurate picture of where the progress toward an OKR might end up at the end of a quarter. This is obviously hard to achieve with complete certainty but even when accounting for the usual margin of error, it's still a valuable way to indicate to everyone which OKRs are on track and which ones are struggling.
Notes: Along with updating the weekly percentage of the OKRs, each owner can add a few notes about what is happening. This can be particularly valuable when a key result is not moving (aka still at 0%) but the team is making good progress with the build which will be deployed in the future and which will subsequently increase the key result pecentage.
Averages: This is simply an average across all the OKRs for the team. If teams are aiming for the usual ~70% achievement rate (aka the OKRs taken together were ambitious but not impossible) this section highlights how the team is tracking on that overall goal.
Team OKR Tracker Examples
The team OKR tracker below has been populated with examples of potential objectives, key results, owners and associated weekly progress tracking for a fictional product engineering team:
Company "Roll Up" OKR Tracker
Each team within the company (e.g. Sales, Product, Marketing, People, etc) has their own OKR tracking sheet which is then "rolled up" into an overall high-level view. This view shows progress across the company against the previously-defined OKRs. It is able to give the executive team, the board and/or the rest of the company (via a monthly all hands meeting or weekly email) a simple view of which teams are doing well and which teams may need some additional help or resources in order to achieve their goals before the OKR period comes to an end.
On January 24th, 2020 Appear Here held its Annual Global Gathering where we brought together world class investors, retailers, landlords and technologists for a day of insights, lessons and knowledge sharing:
My chat with Benedict Evans
I had the pleasure of hosting a fireside chat with the infamous Benedict Evans (formerly a Partner at a16z and now based in London). Benedict is widely considered to be a thought leader for the tech industry and his weekly newsletter frequently shapes the conversation around new tech developments, artificial intelligence, regulation, big tech companies, mobile and venture capital for a global audience. He has an avid following which includes 130k newsletter subscribers and 280k Twitter followers.
We covered a range of topics including high street retail, machine learning/data science and acquisition channels for brands. Below I've included the list of questions we worked our way through during the 30min conversation:
1. You've said. "Everything the Internet has done to media, it's now doing to retail", tell us more.
2. Do you think the high street is dying? Why or why not?
3. What does the future for the high street look like?
4. If it was up to you, in what ways would you like to see retail change or be disrupted by tech?
5. What role do you think machine learning/data science could play in retail?
6. We hear a lot about machine learning and data science, what are some useful ways to think about machine learning? (You've said, "Machine learning is the new database" tell us more)
7. As the cost of online advertising increases, how do marketers continue to grow their businesses?
8. Why do you think Amazon hasn't been as successful with offline retail as they are with online?
9. What do you make of the rise and fall of WeWork; what can we learn from it?
10. After working in SF and London, what differences do you see and what can we learn from them?
Benedict's "Standing on the Shoulders of Giants" Presentation from Davos
Just earlier that week, Benedict had been in Davos to present an annual presentation on tech trends. In a recent blog post he said "Every year, I produce a big presentation digging into macro and strategic trends in the tech industry. This year, ‘Standing on the shoulders of giants’ looks at what it means that 4bn people have a smartphone; we connected everyone, and now we wonder what the Next Big Thing is, but meanwhile, connecting everyone means we connected all the problems. Tech is becoming a regulated industry, but we don’t really know what that will mean." I've pulled out some notable slides from that presentation below that relate to the discussion we had at Appear Here's Global Gathering:
Benedict has clearly thought carefully about how technology is evolving and it was a pleasure to speak with him about those trends as they apply to retail, data science and startups.
