Innovations at Domino’s with Snowflake – Bianca Gilchrist – Domino’s

Bianca Gilchrist, Domino’s Data Literacy & Visualisation Lead and Snowflake Data Hero of the Year, spoke about building a data literacy culture at Domino’s and driving the rollout of Snowflake CoWork, branded internally as Daria, to non-technical staff.

• “Every story, every report that we put out has to have a story behind it” – reports should read like an analyst’s narrative, not a grid of numbers
• Run weekly “15-minute bites” sessions where staff watch live Q&A with Daria on calls – showing how easy it is to ask a question and get an answer, which then prompts a flow of similar questions from the team over following days
• Wrote two data literacy books for children: “the A to Z of data“ and “The Australian Data Story“, and is developing a schools program with Snowflake to bring young kids into the data field
• Three pillars of data literacy: accessibility (can people find the data), usability (what can they do with it), and storytelling
• Marketing data internally – relentless, visible promotion (Viva posts, being put in front of every group) is essential to getting data trusted and used
• One of the biggest recurring business questions the agent gets used for: when a limited-time offer or new product launches, is it cannibalising sales from other products – this previously took a Business Analyst half a day to a week to answer in Power BI, now answered by Daria in minutes
• You need to show a working model quickly, since people can’t visualise an idea from a description alone
• Had to explicitly instruct Daria to be more pessimistic — it was originally overly optimistic (e.g. overstating cost savings), so instructions were added to correct this and avoid dollar-figure recommendations, and to only make recommendations in the voucher area, not financial matters
• Staff are given daily prompt limits partly to control cost, but also to keep people using the core “gold standard” Power BI reports for standard metrics, ensuring the numbers people see match what the CEO sees
• Cost and usage limits are a live tension, staff loved unlimited prompts for two months and pushed back hard on new daily caps
• Future focus is to get the AI agent to redirect simple, reportable questions back to the core Power BI reports, and expanding into specialised sub-agents

Power BI Report development with Snowflake CoCo – Brad de Bruyns – Vivanti

Brad de Bruyns, Senior Consultant at Vivanti (and a former Domino’s colleague of Bianca’s), demonstrated a code-first approach to Power BI development using Snowflake CoCo, showing how AI can turn report-building into a version-controlled, code-based workflow rather than manual clicking in binary files.

• The traditional Power BI PBIX format is opaque and can’t be diffed, whereas the newer PBIP format exposes reports as text so it is version-controllable, diffable, and peer-reviewable in Git
• Snowflake Semantic Views were used as the single source of truth for metrics, shared across both CoWork (formerly Snowflake Intelligence) and Power BI
• Demonstrated CoCo “Skills” baked directly into the repository – encoding branding, DAX standards, layouts, and complex JSON so AI-generated reports and measures come out brand-compliant and consistent every time
• Walked through a live OpenSpec demo: a plain-English request (“add a detailed credit consumption by type page”) was turned into a structured Propose > Explore > Apply process, generating a proposal, design doc, and task list before any changes were made
• Deployment ran through a disciplined Azure DevOps CI/CD pipeline, validating and deploying automatically across Dev, Test and Production
• Underlying agent is CoCo, essentially Claude-like – any Claude-style skills apply the same way
• The Power BI semantic model and the Snowflake semantic view are currently separate but kept in sync via the same underlying spec with a further skill able to uplift a Power BI model into a semantic view
• Manage multiple reports / workspaces in one repo (config-driven via subfolders and deployment targets)
• Started from screenshots / Figma mockups instead of specs and outputs remain fully manually editable afterwards, since it’s still a standard Power BI file underneath
• Less time memorising palettes and file structures, more time on the data modelling and insights that actually matter

Highlights & Key Takeaways from Snowflake Summit 2026 – Majid Miri & Yuqi Zheng – Snowflake

Majid Miri (Senior Solution Engineer) and Yuqi Zheng (Associate Solution Engineer) closed the evening with a rundown of Snowflake Summit 26 announcements, framed around the theme of the “agentic enterprise” and what’s needed to scale AI safely across an organisation.

• CoWork and CoCo rebrands: Snowflake Intelligence is now CoWork, the agent for business users; Cortex Code is now CoCo, the builder’s agent — both reaching a wider surface (desktop, mobile, Slack) since launch
• Horizon Context and Cortex Sense: supply governed business context and definitions to agents, aiming to eliminate the cold start problem of enterprise AI needing weeks of manual semantic modelling
• AI Agent Identity: gives every agent a verifiable, cryptographic identity and a complete audit trail, distinguishing agent-driven queries from human ones for access control purposes
• Apache Iceberg v3: reached general availability, with broad support for new data types, cross-system change tracking, and high performance on semi-structured data paired with Snowflake-managed Iceberg storage to cut operational overhead
• OpenFlow and Datastream: managed data ingestion, OpenFlow for extract/load workflows with new connectors, Datastream as a new Kafka-compatible managed streaming service with separated storage/compute scaling
• Control plane for a governed agentic enterprise – the connective layer of data, context, identity, and governance needed to run AI safely at scale
• Gartner predicts that by 2028 the average Fortune 500 company will have over 150,000 agents, underscoring the need for centralised governance and observability
• Deepened partnership with Anthropic, with Claude positioned as the default reasoning layer underneath both CoWork and CoCo

With 73% of the world using Scrum as their predominant Agile method, which has a direct impact on service management, this session will open up your eyes to the many other Agile and edgy Agile methods and movements in the world today. For many, Agile is a toolbox of potential methods, practices and techniques, and like any good toolbox it is often more about using the right tool for the problem that will result in meaningful results. You may also be surprised about how many methods have a direct relation or reliance on service management as well as the wider organisational structure and culture. So let’s take a rapid journey into the world of methods like Mikado, Nonban, Vanguard and movements like Holocracy, Drive and Stoos where we will uncover 40 methods and movements in 40 minutes to help strengthen your understanding and toolbox.

It was an honour to be invited to Darwin to present this talk to the Darwin tech community who are a small but extremely passionate community. Here are some photos:

She provided a copy of her slides, and here are my notes from the evening:

• Every leader at eBay (440 of them) are Agile Coaches, it’s the third round now for them, imagine the change if you get frozen middle on board
• We need to thrive through uncertainity
• Elon Musk practices first principles ways of thinking

• Compasses are what we need to thrive on uncertainity, we cannot leverage maps because it is an unknown future
• Don’t do more with less, do less, to execute in uncertainity
• Change is changing, we need time for learning and innovating
• If you demonstrate belief in the team and give an environment of safety, the team will believe in their potential – stop telling teams what to do, ask them what they think what we should do
• Safe to fail is critical – we were all born with an Agile mindset (Carol Dweck) but our work and experiences push us towards a fixed mindset – if people can’t learn and thrive, your transformation will fail – as a coach we need to provide air cover
• Keep timelines short all the time – the size of the iteration accelerates the learning cycle and the faster the learning
• Using David Marquet’s Ladder of Leadership model at Ebay – cards you can download, when your employee says this, you say that
• NeuroLeadership Institute – “Why Organizational Growth Mindset Matters“

• An adaptive framework – believing is seeing at centre, need to see awareness and understand the problem, need to process options through discovery (really short time frame, as for 3 value experiments), taking action (learn by doing not thinking), transform learnings into collective knowledge
• DTA has some great tools around discovery
• There is a cost to value, we won’t do anything that doesn’t have x% value, we need to stop being order takers and become value shapers
• Principles for Navigating the Future  (Joi Ito) from the Media Lab at MIT, doing interesting stuff
• Your organisation is not a machine, you can’t fix it – most organisations are setup to work how they were intended to work
• Innovation, growth and transformation does not happen without tension – learn to identify good and bad tension
• What could we do if we knew we couldn’t fail – embrace that
• Do Stanford free Design School program online – same as the expensive in person program
• Polarity Management – polarity is when you think you nailed a wicked problem and then it comes back to bite you, need to find the best win-win from any scenario, if you try to solve it traditionally you make it worse
•  VUCA is here to stay, learning is our competitive advantage
• Microsoft’s CEO Sent an Extraordinary Email to Employees After They Committed an Epic Fail
•  Measure real value, speed to value and cost of value, need relative value not precision because it doesn’t serve us – Case Study and spreadsheet to calculate value

My notes from the session:

• if you need to give the audience bad news, give it first
• instantly does not mean effortlessly

1. Talk about your passion

• to feel more confident, you need competence – talk about subjects you genuinely understand
• seeing someone who is excited… is exciting
• energy, enthusiasm and passion through your actions and speech will translate to your audience
• find something in the required topic that gives you passion – even if you loathe the topic or have been forced by your boss to present it

2. Tell them a story

• our memory is very volatile – stays for 8-10 seconds unless we do something with it
• 7+/- 2 is horribly optimistic and not backed by research, real number is 4 +/- 1
• stories are our oldest information processing tool
• stories have a flow to assist acquisition and memorisation (all our memories are reconstructed from a storyline), have a hierarchy to assist comprehension and recollection
• tell the historical story or the story of what happened, process or funny anecdotes
• story is for your benefit to get the sequence and content right – audience don’t necessarily need to know
• stories make complexity comprehensible, structure recognisable, information easy to remember, make audiences feel more comfortable

3. Don’t search for content, select it

• what should I say is the wrong question, question you should start with is what could I say
• humans are good at recognising important stuff rather than recalling important stuff
• start with a blank sheet and write down everything you know about the topic that you might want to say – stream of consciousness
• whittle down to 3-5 most relevant and important topics to talk about
• these 5 points becomes the chapters, so go looking for the narrative that connects them – they may not connect so look for a couple of lesser topics that better connect the 5 important things
• competency – think about the questions you were asking when you were learning

4. Simplify your slides

• tools encourage a bad job
• content matters but not as much as style
• content is your payload to explode the audiences brains
• style – the stuff the audience doesn’t see that prevents them seeing what they should see
• bad style – anything that prevents the audience seeing what they should see
• a wall of text – technical audience will read everything, regardless of whether it is relevant or not
• Apple is good at presentations – simple but effective
• big words – people at the back can still read them
• slide numbers turn your presentation into a death march – get rid of background, name and title on every slide, get rid of the logos (audience sees salesperson)
• slide deck is to focus audience on the presentation – if they need context give them a separate PDF or notes
• each message is a different slide
• cluttered is overwhelming and as a result they switch off the attention channel as they are trying to read everything
• show less on more slides

5. Manage the questions

• a presentation should always be for the benefit of the audience – give them what they need
• have an explicit questions policy – hold to the end of each topic, end of the talk, or interactive through the talk (can however affect the flow)
• always be keen to take questions – shows you care
• make the questions fit in with your question – “that’s a really good question” makes others more comfortable to ask question

6. Animate code simulations

• explain code temporally, not spatially
• use animations to reveal information one thing at a time
• walk through code as an animation and highlighting
• low tech animations – use the same slide over and over – cell animation
• don’t export your slides – notes
• live coding – synthesise, automate or have a partner – need to keep contact with the audience

7. Deliver your message fearlessly

• use your nervousness – turn fear into energy
• never give a presentation for the first time – practice it live at least 3 times
• use an audience image on a big screen

• used Kanban – work item types to allocate capacity and 3 levels of service (standard, coordinated, expedite)
• JIRA was the tool but was not the most ideal choice and not really up to the task
• reported by type on backlog, work in progress, throughput (number of work items not size), velocity (throughput and velocity were similar which showed average size), cycle time, class of service mix, due date performance, estimation accuracy, cancelled WIP (started and then cancelled work) and demand balancing (clearing out the backlog)
• flow efficiency – percentage of time you work on an item versus in progress, also referred to as touch time – using time tracking in JIRA to do this
• time allocation – value add, failure load (defects), transaction cost (overheads of planning and releasing), coordination cost (management), used percentage of time spent rather than actual hours

The one things that got me thinking during this presentation was the flow efficiency report.

Early on my journey of being an Iteration Manager, my teams used to track times per card. We used to use XPlanner which had some reasoanbly easy functionality for tracking time (one of the good features was as the Iteration Manager I could enter time for my team if needed, tools like JIRA require the developer to record that data if you want it assigned to that developer). We used to use thee metrics for comparing estimates to actuals but over time I came to the conclusion that we would be much better off just making sure that the cards were completing on time (an average of 3 times) and splitting cards out if they appeared to big.

Lately, a number of people in my Agile classes have been arguing that time track is beneficial. My usual response to this (like for all metrics) that it is OK if it adds value, but my recommendation is not to waste your time. Even more so, this opens up the estimation debate that I also believe that a lot of time should not be wasted on (#noestimates), but that is a discussion for another post. My main reasoning is often we need to track for other sources (like timesheets) in different systems, and the overhead does not justify the effort. If teams need time metrics (often to see if time is being wasted away from the core work of the team, say on production support or corporate meetings), I suggest they are done at a team level and rounded to the nearest hour, and collected as time not spent not on project work.

In the graph above, flow efficiency is a good way for showing waste in the system (in this example, the team could potentially be way more effcient), but it relied on the team tracking time (in this case using the time tracking feature in JIRA against each card). I really like it as a graph, I am just not sure the effort to produce it is justified.

Some discussion in the Q&A revolved around recording time tracking (or similar metrics) is OK if the team understands it is an incentive for better metrics, and I can’t disagree with that thought. Just in my experience as an Iteration Manager, getting reliable and timely time and effort metrics ha been painful and the reward outweighed the effort.

Here were a couple of items I picked up on his talk on Ember.js:

• like Ruby and Rails, Ember.js is not optimised for computer science but for developer happiness
• ambitious app – optimised for building entire applications, long lived applications (like Gmail, lasts on the screen for a very long time) need a sophisticated architecture
• MVC is not quite enough – model (all sessions), controller (just this session), view (what is currently visible)
• use HTML and CSS to layout application, is the lingua franca of the web
• HTML needs more dynamic features – {{ for data binding, # block form in Ember
• URLs are important – good applications have good URL support – like Twitter or Rdio – primary way that people share information in Twitter and email for example – is the UI of the web, we should use them more universally and use them everywhere

Yehuda then gave a demo. Having not used Ember.js before I was impressed by how easy it was to build URLs and interactivity. I certainly will give it a closer look at some stage.

• future of the web – take a look at the google project – Polymer
• evergreen browsers – Chrome started this, once you download the browser that is permission enough to download new versions of the browsers ; Safari has not adopted this and will probably become the next IE6!
• some good resources include embercasts and Ember 101

There was also an interesting kick off presentation by Toby Hede about some of the cool stuff Ninefold are doing, if for nothing else the animations were amusing!