<![CDATA[The Civic AI Observatory]]>https://civicai.ukhttps://substackcdn.com/image/fetch/$s_!V5ug!,w_256,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F94c0bafc-f775-4152-bce4-48c194e2b2b9_400x400.pngThe Civic AI Observatoryhttps://civicai.ukSubstackMon, 13 Apr 2026 19:45:29 GMT<![CDATA[Great Expectations]]>https://civicai.uk/p/capabilitieshttps://civicai.uk/p/capabilitiesWed, 19 Jun 2024 15:39:58 GMTThe Civic AI Observatory (civicai.uk) is a new initiative from Nesta and Newspeak House to support organisations working for the public good as they plan and adapt to the rapidly evolving field of generative artificial intelligence. Resources, case studies, events, online spaces, and more.

  • Doing anything with generative AI? Seen something useful? Tell us about it: hello@civicai.uk

Great Expectations

There are now major new AI language models being released every few weeks. They're gradually getting better, both in "intelligence" (AKA "capability"), and also in "multimodality" (the ability to work with multimedia inputs and outputs). A dozen or more companies are releasing these models, and there's no particular model that significantly surpasses the others; at the moment, choosing between them is an engineering decision. Open source models are more or less keeping pace with proprietary models, largely thanks to Meta's Llama series of open-source models. There are also now many smaller models being developed specifically to run locally on phone hardware.

Investment and attention in this sector is being driven by the belief that model capabilities will continue to improve, leading to significant new applications appearing in the medium term (6-18 months), such as reliable autonomous agents. However, there's no guarantee that the current rate can be sustained – creating them is becoming extremely expensive, all easily available training data has been used, and further improvement will only come either through breakthroughs in model architecture, or innovations in synthetic data for model training.

Language models have been put to various applications, which broadly can be characterised as text extraction, summarisation, classification, translation, editing, drafting, ideation, and software development. The models still make mistakes, and so most organisations are choosing to either manually supervise the outputs, or apply them in settings where perfect accuracy isn’t required, such as statistical corpus analysis. An example of the latter can be seen in this case study where AI was used to analyse hundreds of thousands of lengthy Financial Ombudsman decisions.

We don’t consider these applications to have significant strategic implications for most organisations at this time. It remains prudent to wait for the product landscape to mature rather than to make investments – the software packages that are already in use in your organisation will continue to get AI-enhanced features. For example, both of the Google and Microsoft enterprise suites are rolling out their own AI customer service systems.

On the other hand, there are strategic implications from the growing risk of sudden and unpredictable changes in user behaviour. There are two big developments coming that will mean many more people are interacting with AI in everyday life. Firstly, most smartphones will soon have integrations built into the operating system. Secondly, Meta has announced that they are rolling out AI features in WhatsApp, Instagram and Facebook Messenger. It’s not clear what these will mean in practice; one possibility is that AI becomes a significant alternative for search, which may affect your organisation if it depends a lot on search results, for example if your users come directly to you for legal aid or medical advice.

Up to now we’ve only seen two unmistakable changes to the way the world works: how students write essays, and how jobseekers write applications. It’s notable that these are both “interim” activities where process automation is intentionally suppressed. This fits the pattern: so far, mature existing systems are still better than what these models are capable of on their own, and the problems arise with those who have to deal with the downstream consequences.

Observations

  • AI risk management has become a thriving ecosystem with hundreds of companies creating products to tackle issues with governance, compliance, quality, security, and privacy that arise when putting AI models into practice. link

  • A prototype tool that extracts the structure of a form from an image or PDF and generates a multi-page web form in the GOV.UK schema. link

  • National Audit Office report on UK government AI adoption strategy: “Our survey of government bodies found that AI was not yet widely used across government, but 70% of respondents were piloting and planning AI use cases.” link

  • Policy and Guidance for AI in UK government procurement: “Planning for a general increase in activity as suppliers may use AI to streamline or automate their processes and improve their bid writing capability and capacity leading to an increase in clarification questions and tender responses.” link

  • CAST launches AI resources hub, including communities for digital leads and grantmakers, as well as grants available for experiments. link

  • A collection of case studies of civic services delivered via AI-powered WhatsApp Chatbots. link

  • Can we protect elections from artificial intelligence? link

  • How generative AI chatbots respond when asked for the latest news. link

  • The Impact and Opportunities of Generative AI in Fact-Checking link

  • International Fact-Checking Network renews its grant program for leveraging WhatsApp to track or respond to AI-generated misinformation. Deadline 8 July! link

  • BBC Editorial Guidelines on the use of Artificial Intelligence: “Generative AI should not be used to directly create news content published or broadcast by BBC News…” link

  • UK Department of Education report on Generative AI in Education: “In addition to augmenting educator jobs and tasks, GenAI could also fundamentally alter how and what people learn by changing how information is synthesised and presented.” link

  • Contribute to a survey on the impact of AI on knowledge consumption and production processes, specifically research, cultural heritage and education. link

  • Joseph Rowntree Foundation report on the implications of AI for digital inclusion: “People started talking about how AI could help them, and how using AI was something they could do too.” link

  • Mozilla launches the AI Intersections Database, which maps key social justice and human rights impacts as well as entities working on them. link

  • National Cyber Security Centre assessment of the impact of AI on cybersecurity over the next two years: “All types of cyber threat actor – state and non-state, skilled and less skilled – are already using AI, to varying degrees. AI will almost certainly make cyber attacks against the UK more impactful…” link

  • Digital Regulation Cooperation Forum launches AI and Digital Hub, where innovators can obtain answers to complex questions that span the remits of the CMA, FCA, ICO, and Ofcom. link

  • Stanford AI Index Report covers technical advancements, public perceptions, geopolitical dynamics, estimates of model training costs, analyses of the responsible AI landscape, and impact on science & medicine. link

  • 100 Days of AI: a series of a hundred 30 minute exercises to improve your AI skills. link

Next steps

  • Doing anything with generative AI? Seen something useful? Tell us about it: hello@civicai.uk

]]><![CDATA[Safe Bets]]>https://civicai.uk/p/safe-betshttps://civicai.uk/p/safe-betsWed, 03 Apr 2024 09:56:00 GMTThe Civic AI Observatory (civicai.uk) is a new initiative from Nesta and Newspeak House to support organisations working for the public good as they plan and adapt to the rapidly evolving field of generative artificial intelligence. Resources, case studies, events, online spaces, and more.

  • Doing anything with generative AI? Seen something useful? Tell us about it: hello@civicai.uk

Safe Bets

Despite what many are keen to tell you, developing a strategy for generative AI is difficult. It is such a general purpose technology that it can be applied to almost any process in your organisation. Outside of generic use cases like the ones covered in the last issue, experimentation requires expensive skills. The technology also remains frustratingly unpredictable, a significant concern for civic organisations that must be extra careful to mitigate any risks to users, and also to maintain public trust. Despite these challenges, the impressive capabilities and rapid improvement of generative AI makes it hard to ignore. So if you’re a larger organisation that has some capacity to start exploring, how should you decide where to invest?

In this issue we are highlighting the work of the UK Cabinet Office’s new Incubator for Artificial Intelligence. While it’s easy to be cynical about such government initiatives, the team is impressive, and the three projects they have announced so far are widely applicable and align closely with our assessment of the current most promising areas for custom applications. We believe these projects are a good indication of realistic civic use cases we’ll be seeing a lot over the next year.

Following and learning from how these projects develop will help other civic organisations develop confidence for trying similar initiatives of their own. As a bonus, the incubator’s emphasis on open source will make it much easier and cheaper to replicate them, and means your organisation is less likely to get stuck maintaining an orphan codebase.

Finally, the fact that central government is actively experimenting with these projects at scale sends a strong signal to the civic sector. If they deem these initiatives as acceptable and worthwhile, it provides a level of assurance and legitimacy for smaller organisations to explore similar applications, to some extent alleviating concerns about the risks and uncertainties associated with this emerging technology.

The three projects they’ve announced so far are the Consultation Analyser, Redbox Copilot, and Caddy, with more promised soon. Here is info about each:

Consultation Analyser

The UK government conducts around 800 public consultations annually, with popular topics getting hundreds of thousands of written responses. Transforming unstructured text into structured data is one of the more conventional and labour-saving applications of large language models (LLMs), and this is an obvious use case.

While more traditional natural language processing techniques remain surprisingly relevant in the age of LLMs, the latter are rapidly closing the gap in terms of performance and are much more versatile, making it possible to tackle a wide range of text analysis tasks with a single, powerful tool.

The Consultation Analyser extracts and summarises themes from consultation responses, and presents them clustered on a dashboard where analysts can drill down to see the raw submissions. It’s been piloted it with the Department of Health and Social Care running in parallel with their existing consultation process, and more tests are now being rolled out to different departments.

For your organisation, this same approach could be used to help analyse any kind of large unstructured text dataset almost instantly and much more cheaply than any manual analysis. This opens up the possibility of, for example, running much larger and more frequent surveys that include more free text fields. It could also be used for inputs gathered via any other channel, such as Whatsapp, email, social media, or even phone transcripts. If your organisation has a large and active membership that have a lot they want to say – to you or to each other – understanding and engaging with them is becoming much more feasible.

Redbox Copilot

After a few months of being trialled in ministers’ offices, Redbox Copilot is now being rolled out to the wider civil service. Named for the iconic red boxes used to transport state papers, this project uses an AI technique called retrieval augment generation (RAG) to search and collate information from the many thousands of documents held across government, allowing civil servants to chat with a clued up ChatGPT-style assistant, or rapidly produce a summary of a set of internal documents complete with citations.

The project has had to tackle all of the compliance constraints that come with public sector deployment such as data privacy and retention, as well as come up with acceptable standards for the AI specific challenges: safety, bias, and hallucination. The project itself, or parts of it, may be useful if your organisation has growing piles of (potentially sensitive) internal knowledge that nobody has the time to think about.

Caddy

After Klarna’s recent announcement that it saved $40m by switching to AI powered customer service, there’s been a lot of interest in how replicable this is and whether it could safely be deployed in the public sector. Caddy, a collaboration between the Incubator for AI and Citizens Advice, is a tool that uses a mix of internal and external information sources to draft responses for human advisors, thereby avoiding most of the risks of unsupervised AI.

Caddy is built to be interoperable and slot into a range of different systems. The prototype is built on Google Chat, with an integration for Microsoft Teams coming soon. It is designed to make it easy to use data from many different sources; the current prototype combines internal Citizens Advice training material, the contents of the public Citizens Advice website, as well as all of GOV.UK.

A randomised controlled trial of Caddy at Citizens Advice centres will start in the coming weeks, looking to understand how it's impacting both advisors and advisees. Recent research suggests this kind of AI support will give a modest improvement in productivity with no loss of quality, and is particularly useful for helping new advisors get up to speed. If the results look good, the tool could be used in many places across government and beyond, and the incubator is actively seeking partners interested in exploring applications.

Next steps

]]><![CDATA[An Infinite Shortage of Managers]]>https://civicai.uk/p/productivityhttps://civicai.uk/p/productivityMon, 26 Feb 2024 17:26:43 GMTThe Civic AI Observatory (civicai.uk) is a new initiative from Nesta and Newspeak House to support organisations working for the public good as they plan and adapt to the rapidly evolving field of generative artificial intelligence. Resources, case studies, events, online spaces, and more.

An Infinite Shortage of Managers

In civic circles, a consensus is emerging that the current large language models are too unreliable to use in serious contexts without supervision. With the present state of the technology, anything produced by a large language model should be reviewed by a human before it’s used (e.g. gov.uk, judiciary). They still confidently get things wrong and are vulnerable to prompt injection and other security risks. While this may be very different in 6 months’ time, ambitions of programmatic deployment at scale remain risky for now.

However, this is not to say that the technology is useless! The main applications are in “increased productivity”: on average employees using AI work significantly faster and better, though it’s hard to predict which tasks will benefit from AI and which ones will not, and the exact technique will be quite particular to the piece of work, like it might be with a custom spreadsheet. The implications of this are surprising: likely there will not be any substantial change in the nature of work being done, nor should you start new initiatives or hire new roles specifically related to AI. However, parts of your organisation may suddenly be able to move faster than before, so you might find that your personnel balance has to change. In particular, roles related to managing risk—anything to do with quality assurance, supervision, review, or getting sign-off—may become overwhelmed, frustrating motivated team members. Tech pundit Benedict Evans famously says AI is like having an infinite supply of interns; from a strategic perspective, what this actually means is having an infinite shortage of managers.

In this issue, we’ll be looking at the generative AI integrations appearing in common enterprise software that your organisation is likely already using. Even if you don’t have a plan to use generative AI in your organisation, it’s coming for you anyway! Most major tech firms have announced a host of new AI integrations into their product lines, launching in various states of usefulness and coherence - however these present an immediate and everyday way that your team will start using generative AI. While the strategic situation may change as the technology and product landscape improve, what we do know is that these tools present immediate opportunities and challenges, and will at least be thought-provoking.

It’s worth spending time reflecting on what uneven changes in productivity might mean for your organisation. Pro-actively discovering and resolving bottlenecks will be critical. Small projects previously shelved may start looking viable. Even modest improvements in the efficiency of very expensive activities—e.g. anything involving lawyers, developers, or senior roles—may well be worth investigating.

If you’ve had any interesting experiences like this, or you strongly disagree with this take, please let us know! hello@civicai.uk

🏢 Enterprise Suites

For many orgs, their main decision here has been whether to shell out for the AI upgrades of their enterprise suites, chiefly Microsoft 365 or Google Workspace.

Microsoft has now fully rolled out Copilot for Microsoft 365, available for an additional (hefty) fee above the basic 365 subscription, and no longer just restricted to large orgs. At the moment this means new functionality for Outlook, Teams, SharePoint, OneDrive, Word, Excel, Powerpoint, Whiteboard, and OneNote, all confusingly called Copilot even though they behave differently in different apps.

Similarly, Google has begun to integrate AI into most features of Google Workspace under its Gemini model, now called Gemini Enterprise (previously known as Bard as the model and Duet AI for workplace apps). This is lagging behind Microsoft somewhat but presumably not for long.

Reviews of these platforms so far are mixed; they’re expensive and generally considered underdeveloped, but improvements are coming out faster than we can cover them, and we can’t help but admire their vision of end-to-end AI. We’ll cover the changes in different apps in following sections.

📧 Email

It’s the opinion of the Civic AI team that speeding up email will be one of the most tangibly impactful applications of genAI for many organisations: it’s a ubiquitous, non-repetitive, piecemeal task requiring lots of context switching, all things for which language models are especially helpful. Microsoft has released Copilot in Outlook and Google has a feature to draft emails with AI, both of which can summarise email chains and draft emails from prompts. Gmail has third party plugins like Shortwave. Superhuman has every feature you could think of: summarising, drafting, autocompletion, editing; it feels like these kinds of enhancements will become standard in future, and products will be differentiated by what other context they can intelligently draw from (calendars, documents, etc).

💻 Calls

Generative AI as applied to videoconferencing is mostly for making notes and summaries of meetings via transcripts. Transcription technology has been improving quickly, and with a quality transcript, LLMs can provide summaries and answer questions about the meeting, even while it’s happening. It works relatively well, though LLMs can miss tone or emphasis, and get confused if a discussion requires a lot of nuance or additional context not included in the call. In short, automated summaries are better than nothing but are best used for a high-level overview only, and don’t yet take notes that are as useful as those taken by a human.

Microsoft’s Copilot in Teams, Google Meet’s AI extension, and Zoom’s AI companion all offer versions of these capabilities. To set up AI in Teams is fiddly: create a meeting, go to additional settings to allow transcription, send the invite, then once the meeting begins, manually start transcription. With luck you should then be able to start Copilot.

Zoom’s AI Companion is worth trying but users have been split on performance: transcription is hit and miss, it doesn’t always understand which points are important, must be activated by the host to work, and is limited to English only. We’ve been recommended fathom.video as an alternative, a Zoom add-on which gives video transcripts, video recording, and AI summarisation.

There’s very little of substance from Google Meet as yet, but updates for Gemini are coming out quickly at time of writing so this may change very soon. Likely similar functionality to the others.

📱Chat

Here we’re referring to where your team sends each other instant messages, not AI Assistants like ChatGPT. It’s no coincidence they look similar though; ChatGPT has leveraged the UX of instant messaging to teach people how to interact with it, an example of skeuomorphism, which can be confusing when it’s put next to things that are actual chats with other humans.

How AI should work with team chat is still being explored: it’s appeared as an integrated AI assistant (Copilot, Discord), a tool to summarise conversations and threads (Slack, Teams), or in the case of Slack, a whole genAI appstore and automation builder. If your team mostly uses chat to communicate, and lots of information comes into the channel throughout the day, features like Slack’s thread summaries can be very helpful. 

For Microsoft enterprise users, the situation is confusing. Chat-like experiences have appeared as both M365 Copilot Chat and just Copilot. M365 Chat appears as another conversation within Teams and designed to appear as if you’re chatting with a colleague while the other option appears through a separate portal and acts as a more familiar AI assistant like ChatGPT. Only M365 Chat has access to your chat history with other team members. It also gives a compose button that helps you write messages, and also a copilot button that summarises chat history.

Google’s Gemini will now have access to Google Messages, an app used by over a billion people. Since this is a very recent development, we don’t know how good it is. However, it’s already caught the attention of EU regulators who may see it as a violation of EU law, which prohibits the use of private messages for LLM training. 

WhatsApp hasn’t announced much on this front, except the ability to use AI to create new stickers. However, we have seen a few third-party approaches, such as the option to integrate ChatGPT into WhatsApp or create a WhatsApp support chatbot using LLMs, OpenAI, and Python.

✨ AI Assistants

For many people AI Assistants like ChatGPT have become just another kind of work tool, and so merit inclusion in this list as their own category. Most civic organisations are already widely using these for ad-hoc tasks, mostly drafting of text.

While Microsoft and Google’s enterprise suites both have AI Assistants built in—Copilot and Gemini (previously Bard) respectively—many prefer or at least are more habituated to ChatGPT, which is furiously building out its B2B offering with ChatGPT Enterprise, including things like file uploads, data analysis, plus all the privacy and procurement infrastructure you’d expect from an enterprise product. Meanwhile in its mission to integrate-all-the-things, Microsoft has also released Azure ChatGPT competing with their own Copilot, which allows you to build integrations with your internal data if you’re in their ecosystem already. If you’ve tried using it to make anything interesting, let us know!

📅 Calendar

Microsoft Copilot can “intelligently” schedule meetings from Outlook, suggesting available times, relevant people and files, and even draft an agenda. Google Workspace hasn’t announced an AI calendar integration but doubtless will in coming weeks. Calendly is also testing AI integrations, currently available via waitlist.

📦 File Storage

One of the big promises of language models is to be able to help organisations leverage their knowledge base. This is still emerging, but new features from Microsoft, Google, and Dropbox into their existing products will likely be the first taste people will have of that. Gemini’s just released integration with Google Drive is impressive and the company has teased more features are coming. Microsoft OneDrive provides recommendations for files based on associated meetings or colleagues. Specific tools for this, such as Glean, have also appeared. Glean’s very decent list of clients suggest that large organisations are experimenting with it in addition to the AI features embedded in their enterprise suite.

📝 Docs & Notes

Both Google Docs and Microsoft Word can provide drafts based on a single prompt or a prompt that references an existing file. For example, a user could type, “Write a memo to stakeholders based on the ‘2023 Cybersecurity Report’ document” and the new feature would mock up a document according to those instructions. Copilot in OneNote similarly is an AI assistant with access to your notes, letting it do things like create summaries, to-do lists, or give ideas and feedback. The note-taking app Notion also lets you write, refine and summarise text with AI. If you’re familiar with ChatGPT these are all more or less what you’d expect; we at the observatory are more excited about clever AI augmentations of writing and editing that are still being invented.

📈 Sheets

Both Microsoft Excel and Google Sheets boast native AI integrations but they’re very rudimentary, and we don’t recommend them. If you’re adept with spreadsheets these won’t be useful, and if you aren’t, then they might make formulas you don’t understand, perhaps leading you to misinterpret the output. In Excel, you can ask it to, for example, “analyze data in a given document and ‘provide three key trends’”. Google Sheets appears a bit more behind, with only the ability to suggest table headings. 

Much more interesting are addons like sheetai.app for Google Sheets and ChatGPT for Excel that let you use language models directly within your spreadsheets as functions. These are more limited than full-on AI assistants as they don’t retain context, and both prompts and responses need to fit within cells, but despite this the flexibility of spreadsheets allow for many creative uses.

🖼️ Slides

Microsoft and Google both offer AI-enhanced slides but differ quite significantly on what they can provide. So far, Google Slides only allows you to use Gemini to create images. With Microsoft Powerpoint, you can create a full draft slidedeck based on a prompt and an existing document. This leaves you with “good bones” for a formal presentation, as one of our interviewees described it. Other tools like gamma.app are also appearing which seem very slick.

🎨 Boards

Microsoft Whiteboard, Miro, and Mural have all introduced similar AI-powered features that suggest new ideas or provide summaries of brainstorm sessions. We have received mixed reviews on these; if you already have templates that you go use regularly you won’t need them, but if you’re making new boards from scratch all the time they might help you get started.

📊 Project Management

Your team may be attached to a certain project management tool, whether that’s Airtable, Trello, Monday, or Asana. Each have announced AI-features that are writing and categorization oriented, though many are still in beta and not widely available. If you’ve tried these and have an opinion on whether they’re useful, let us know.

🙇 Customer Service

AI-enhanced customer service comes in the form of chatbots, or writing assistance for agents. Zendesk, HelpScout, Intercom, Microsoft Contact Centre, and Front have all introduced some version of AI-enhanced features. There are even some creative ways to do things with WhatsApp. These are quite good, though not miles ahead of what customer service chatbots have been able to do before now and can be vulnerable to prompt injection, or at least being made to say silly things. If you’re giving legal advice or helping vulnerable people you should probably steer clear of these, but for less sensitive applications they’re worth exploring.

🔢 Data & CRM

There are all kinds of varied experiments going on in this product area with different levels of sophistication, focus, and customisability; databases are generally integrated with other systems and it’s not very clear at which points of the stack AI functionality ought to sit, so it’s interesting to see how different companies are approaching this. What it will actually mean for you will also depend a lot on how your organisation uses these tools.

Airtable’s AI integration gives you a lot of control, allowing you to create fields that produce an AI response to other fields using a specified prompt. You could do almost anything with this; the obvious use for this is extracting structured data from text, even quite abstract data like sentiment, but usual caveats around model accuracy apply.

HubSpot and Zoho offer an integrated AI assistant that can edit your database for you, as well as generate content drawing from it. If you’re using Microsoft Suite or Salesforce, they claim to allow you to do this directly in your existing apps like your email or customer service channels, which has the potential to be very powerful indeed. Finally CiviCRM hasn’t committed to a product direction yet, but is doing some small experiments to explore the technology.

🎁 Images & Ads

With the proliferation of AI-created images from Mid-Journey, ChatGPT’s DALL-E and Google Gemini, all other creative software packages were quick to offer in-house versions. Canva has launched an AI-powered version of its image design platform, Adobe offers powerful AI-features in Photoshop, and Shutterstock now includes Creative AI, an array of enhanced editing capabilities. Getty Images launched a tool for “commercially safe” generative AI images which claims to exclude known people and copyrighted material.

Both Meta (Facebook) and Google are exploring AI-assisted content generation for their ad platforms. Meta’s will automatically create variations on copy and images for different ad formats, whereas Google Ads will generate a whole campaign based off content from your website. A bit surprising not to see faster innovation here given that these are very much core to both companies, but it’s likely that if you’ve not explored online ads for your campaigns it will soon become extremely easy to do.

🧰 Other

  • Wordpress has a variety of AI-powered plug-ins.

  • Webflow has released various AI features such as generating, altering, translating, and summarising static and CMS content, and AI powered SEO optimisation.

  • Buffer introduced an AI assistant that can generate new posts for a schedule and repurpose existing posts.

  • SurveyMonkey and Typeform are promising AI powered survey building.

  • Zapier and IFTT have introduced AI actions, offering mindbending new possibilities.

  • Quickbooks and Expensify can help categorise and reconcile payments, and check invoices and receipts.

  • Yoti has released new features to make it more resistant to AI generated faces.

Is there other enterprise software you use all the time that we should be thinking about? Tell us what! hello@civicai.uk

Next steps

  • If you’ve had good or bad experiences with any of the software here - tell us about it! hello@civicai.uk

]]><![CDATA[Organisational Policies for Generative AI ]]>https://civicai.uk/p/issue-1-organisational-policieshttps://civicai.uk/p/issue-1-organisational-policiesTue, 03 Oct 2023 12:19:37 GMTThe Civic AI Observatory (civicai.uk) is a new initiative from Nesta and Newspeak House to support civic organisations as they plan and adapt to the rapidly evolving field of generative artificial intelligence. Resources, case studies, events, online spaces, and more.

Your organisation needs some new policies

In this issue we’ll be looking at organisational policies: even if your organisation does nothing at all with generative AI, you will still need them. The widespread accessibility of primary interfaces like ChatGPT, as well as the emerging landscape of generative AI tools, have the potential to affect many different parts of your organisation, with both opportunities and risks. Responding to this will likely require a more general organisation-wide strategy than what you’ve had up to now.

Here are all the different kinds of organisational policies that we’ve seen:

1/ Internal policies for staff use

First up - people in your organisation will already be using it, so you’ll want a policy on employee use. This will bear a passing resemblance to a “use of social media” policy. Individuals and teams across organisations are already actively experimenting with tools such as ChatGPT, and it’s imperative to acknowledge this and introduce appropriate structures to steer such experimentation in the right direction.

This should cover both opportunities and risks. Employees may benefit from training or allocated time to share and explore AI tools together. Depending on your organisation there may also be pressing issues related to security, privacy, reputational risk, or wider ethical concerns. In general, the policies we have seen have been quite balanced, both recognising the possibilities of the technology as well as warning against the dangers.

Here are a selection which may be useful in forming your own policies:

  • The official guidance for how civil servants should use this technology from the UK government 

Let us know if you have one already, or are planning to make one: hello@civicai.uk

2/ Internal policies for procuring generative AI products

It is also important to think about having a policy for any tools that you’re buying. We expect that many agencies and startups will be selling new AI products so it’s helpful to know how to evaluate them and work out which ones make sense for your organisation. 

Here are the best resources we have found so far:

3/ Public statements about the organisation’s use of generative AI

Your customers, members, donors, or other stakeholders may be asking questions about your approach to AI. For many orgs, just publishing the internal policies may be enough. However, in certain sectors such as journalism, AI poses added reputational risks so they are particularly keen to reassure their audiences. Here are samples from various news orgs we have seen:

Despite a few negative stories, public perception of generative AI use in civic contexts has been surprisingly positive. Some recent research from the Centre for Data Ethics and Innovation tells us that most people are open to the use of foundation models within the public sector, especially in low-risk use cases, provided that there’s human review of outputs and accountability over decisions.

If you have seen or done any other research like this, even anecdotal, let us know: hello@civicai.uk

4/ Internal policies for using generative AI in digital services

There’s no doubt that generative AI will be used to improve digital services and enable new kinds of user experience, but this is very new and how it will look is not yet clear. Working this out will need both technical knowledge of what the models can do and detailed internal knowledge of the organisation and service. Digital product teams can start thinking about simple applications, but model capabilities are still improving quickly and effective design patterns will emerge as the ecosystem matures and tooling improves.

We will be tracking this in future issues, but in the meantime, if you are planning a project please tell us: hello@civicai.uk

If your organisation has already been using standard machine learning then perhaps you already have some kind of responsible AI policy, but in light of generative models, you may want to update it. There will be many new possible applications that you may choose to avoid or pursue, and you may want to review your approach to disclosure, testing, data privacy, oversight, and so on. “Non-generative AI” policies are a relatively mature area now - this paper gathers and compares many examples - but we haven’t yet seen many examples that account for generative AI specifically:

Note that we’re not talking about using generative AI to help with software development, although there is an entire revolution going on in this area with tools like GitHub Copilot (these should also be covered in your individual use policy!)

5/ Policies for written submissions

Anyone who receives lots of text-heavy documents - for example, grant applications, tenders, job applications, even student essays - will have noticed that their lives have changed in 2023: many more submissions, often of dubious quality. There’s a whole subfield in using AI to evaluate applications - maybe adding bias, maybe removing it - but in the meantime, you might want to provide guidelines for your applicants as to how or if they should use AI to produce their submissions. We’ve only found one example so far, from The Research Funders Policy Group:

As for education, there’s lots to say here and we’ll likely come back to this in a future issue, but the speed of change is fascinating and might inspire ideas for the emerging problem of written submissions more generally:

A few other interesting things on the governance of generative AI

  • Great research from Harvard Law School surveying governance of AI in corporate organisations

  • Microsoft offers various kinds of governance support for organisations using their AI products 

  • credo.ai is a tool for tracking whether AI systems deployed in an org comply with policies. Perhaps you could do this with a spreadsheet, but it’s interesting to see that there’s now a dedicated tool for this nevertheless

  • Wikipedia policy for editors on LLM use

Next steps

  • For the next newsletter, we’ll be thinking about common enterprise tools that are getting generative integrations. Have any of the tools that you use every day suddenly got some magic new buttons? Are any of them useful? Tell us! hello@civicai.uk

  • Chat with us & other readers about this issue: join our whatsapp group

  • Join us in person at our first Civic AI Unconference! 20th October, 2-5pm, Nesta HQ

]]><![CDATA[Announcing the Civic AI Observatory]]>https://civicai.uk/p/coming-soonhttps://civicai.uk/p/coming-soonTue, 13 Jun 2023 09:58:12 GMT

Nesta and Newspeak House announce the Civic AI Observatory (civicai.uk), an initiative to support civic organisations plan and adapt to the rapidly evolving field of Generative AI.

The Observatory will focus on immediate practical insights on AI for digital leads, including case studies of civic applications, examples of digital strategies and organisational policies, and the impact on the job market, as well as helping to develop communities of practice for specific domains.

Subscribe now

“It is critical that civil society organisations not only keep up with this swiftly changing field but do so in a way that reflects their values. There is immense potential to advance how data is used in public services or social movements and we’re excited to help organisations achieve that through the Civic AI Observatory.”

James Plunkett, Chief Practices Officer, Nesta

While opportunities and risks from AI have dominated the discourse in the past year, it remains unclear how most organisations should develop a realistic strategy for Generative AI. Civic organisations may find it especially difficult due to the challenges they face from limited capacity and resources, and it is tempting to pretend that it’s something that can be safely ignored.

“The implications for civic organisations are not yet obvious and rightly there’s a lot of scepticism, but it’s clear that one way or another there will be significant implications for everyone. This is not just about ChatGPT - soon enough all kinds of digital products will have integrations with Generative AI, including ones your organisation already uses, or that your key stakeholders use! If you’re a digital lead and you want to have a serious answer for how it will affect your organisation, it can feel like a full time job to keep up with what is happening week to week. The Civic AI Observatory is here to help you stay on top of what is relevant to you.”

Edward Saperia, Dean of Newspeak House

You can follow and contribute to the Observatory’s work by subscribing to its newsletter. Through this you’ll get regular insights and analysis, as well as invitations to events and workshops tailored to diverse groups from civil society, including senior leaders, funders and capability-builders, technologists, and practitioners.

Subscribe now

If you or your organisation are currently considering a project related to integrating generative AI into your work, or if you've come across anything interesting in this field that we should know about, we encourage you to reach out: hello@civicai.uk

Advisory Board

An advisory board consisting of a wide range of civil leaders and technologists will contribute to the initiative, including:

  • Lisa Murphy – Technology Lead for the Data for Science & Health team, Wellcome Trust

  • Gaia Marcus – Deputy director of strategy, programme and partnerships, Spatial Data Unit, DLUHC

  • Jo Kerr – Director of Impact & Innovation, Turn2us

  • Francis Irving – Co-founder, mySociety

  • Zoe Amar – Director, Zoe Amar Digital & Charity Digital Skills Report

  • Imeh Akpan – Head of Design Research, TPXImpact

  • Karlis Kanders – Senior Data Foresight Lead, Nesta

  • Jess Morley – PhD in AI & health, Oxford Internet Institute

  • Richard Pope – Co-founder, UK Government Digital Service, Director pt2.works

  • Andreas Kirsch – PhD in Machine Learning, University of Oxford

  • Hannah O’Rourke – Director, Campaign Lab

  • Eirini Malliaraki – Head of Programme, Deep Science Ventures

  • Dan Kwiatkowski – Consultant AI Engineer

  • Kim Lawrie – Head of Creative & Emerging Tech, House 337/Next 15

  • Peyman Owladi – Charity CTO & Consultant

  • Lewis Westbury – Senior Developer, Centre for Collective Intelligence Design, Nesta

  • Matt Stempeck – Curator, Civic Tech Field Guide

  • Maggie Appleton – Product Design, Ought

About Newspeak House

Newspeak House, The London College of Political Technology, is an independent residential college founded in 2015. Our mission is to study, nurture and inspire emerging communities of practice across civil society and the public sector in the UK.

For the past decade Newspeak House has served as the primary hub in London for applied technology and society activity, bringing together a wide array of institutions from academia, government, civil society groups, and tech communities in its centrally located campus. It is also home to a one year residential programme on political technology for emerging leaders in the field. Find out more at newspeak.house

About Nesta

Nesta is the UK's innovation agency for social good. We design, test and scale solutions to society's biggest problems. Our three missions are to give every child a fair start, help people live healthy lives, and create a sustainable future where the economy works for both people and the planet.

For over 20 years, we have worked to support, encourage and inspire innovation. We work in three roles: as an innovation partner working with frontline organisations to design and test new solutions, as a venture builder supporting new and early stage businesses, and as a system shaper creating the conditions for innovation. Harnessing the rigour of science and the creativity of design, we work relentlessly to change millions of lives for the better. Find out more at nesta.org.uk.

]]>