DES303 Week 8: From Focus Detection to Task-Aware AI Senior

Introduction

Week 8 became a major pivot point in my DES303 project. In Week 7, I tested Focus Integrity through the TriScore model and Research Tick / Replay system. That experiment showed me that Ticker could collect behavioural evidence, compare models, and judge focus states, but the system still lacked the most important thing: task understanding.

The problem was not only technical accuracy. The problem was meaning. Chrome could be distraction or research. A static screen could be inactivity or reading. YouTube could be entertainment or a tutorial. Low typing could be avoidance or thinking. More behavioural data did not automatically create better judgement.

Because of that, Week 8 shifted the project from an Artificial Intelligence (AI) focus detector into a task-aware AI senior. Instead of only asking, “Is the user focused?”, I started asking, “What does the system need to understand about the task before it has the right to judge the user?”

This week therefore focused on building a broader Business-to-Business (B2B) workflow. I explored how Ticker could take a vague work brief, ask clarification questions, generate a project plan, assign tasks, create Standard Operating Procedures (SOPs), and then use that structured task context to support Focus Integrity.

This made the project stronger because Focus Integrity was no longer judging behaviour by itself. It was judging behaviour against a clearer task contract. However, it also made the project more ethically uncomfortable. If AI defines the work, assigns the worker, writes the process, and checks whether the worker follows it, then AI is no longer only a tool. It begins to act like a senior, manager, or authority figure.

Evidence of this week's experiment

To make Week 8 more evidential, I documented the pivot as a design experiment rather than only a conceptual change. The aim was to test whether task context could make Focus Integrity more meaningful.

Experience

Why I moved from detector to AI senior

My Week 7 experiment showed that Focus Integrity could not begin with surveillance. The TriScore model and Research Tick model were useful because they helped me test different evidence strategies, but they also exposed the limit of my method. I was mainly trying to improve the system by adding more signals.

This created a problem. The system could become better at watching the user, but not necessarily better at understanding the user. The more evidence I added, the more technically complex the system became. But if the task was still vague, the judgement was still weak.

For Week 8, I therefore changed the starting point. Instead of asking how to collect more focus evidence during a session, I asked how Ticker could understand the task before the session starts.

This led to the idea of the AI senior. The AI senior is not only a chatbot or timer assistant. It is a system that helps turn unclear work into structured task context. It can ask questions, define outputs, break work into phases, assign responsibilities, and create SOPs that later become useful for Focus Integrity.

What I built: B2B AI senior workflow

The main system I explored this week was a B2B AI senior workflow. This moved Ticker away from being only an individual productivity app and toward a future workplace system.

Briefing → AI clarification loop → AI plan generation → Review and edit → Auto-delegation → SOP generation → Final approval → Worker handoff → Focus Integrity

This workflow mattered because each stage created more task context. A vague brief became clearer through AI questions. The clarified brief became a plan. The plan became tasks. The tasks were assigned to workers. The assigned tasks became SOPs. The SOPs then became focus contracts that Ticker could use during a focus session.

This made Focus Integrity more believable. Instead of judging the user from a rough task title like “work on project”, Ticker could judge against a more specific contract: what the user is meant to produce, what tools are expected, what behaviours are acceptable, and what would count as drift.

B2B SaaS website proof

I also built the Week 8 B2B SaaS direction as interface screens. These screenshots are important evidence because they show the AI senior workflow becoming a product surface: creating a project, adding people, defining phases, creating tasks, and viewing delegation against a timeline.

B2B SAAS PROOF · HOME DASHBOARD SCREENS

Open larger viewOpen larger view

B2B SAAS PROOF · PROJECT SETUP FLOW

Open larger viewOpen larger viewOpen larger viewOpen larger view

B2B SAAS PROOF · FULL WORKSPACE WITH DELEGATION

Open larger view

Worked example: turning a vague brief into a focus contract

To make the Week 8 experiment more concrete, I tested the workflow through one example task. The vague brief was:

Create a landing page for Ticker's B2B version.

On its own, this task is too vague for Focus Integrity. If a worker opens Figma, Chrome, YouTube, Notion, or a document editor, the system cannot easily know whether that behaviour is aligned or not. The task needs more context before it can be judged.

The AI senior can then turn this into delegation logic. That delegation depends on role, skill, workload, availability, and task requirements.

Finally, the AI senior can turn the task into a focus contract.

This example helped me see why task context matters. Focus Integrity becomes more useful when it can compare behaviour against a clear task contract instead of judging activity by itself. This also showed that the AI senior needs clear task specification before it creates SOPs, and clear evaluation criteria before Focus Integrity checks the work.

Ideation: How could Ticker be experienced at capstone?

After Week 7, I knew that Focus Integrity needed richer task context. At first, I started solving this as a software problem: AI planning, task delegation, worker rosters, and Standard Operating Procedures (SOPs). However, during Week 8, I realised I also needed to think about the final capstone format. This project is not only a Business-to-Business (B2B) product idea. It needs to show a future condition clearly enough for people to question it.

Because of that, I started brainstorming different ways Ticker could be experienced, not just used. This was important because a normal dashboard might explain the system, but it may not make people feel the pressure of being watched by an AI senior.

This ideation stage helped me realise that the strongest capstone direction is not necessarily the cleanest product demo. The strongest direction is the one that makes the future most visible. The workplace simulation became the strongest option because it lets the audience enter the system, receive a task, be monitored, and feel the pressure of being judged by AI.

Converging: Why the workplace simulation became the strongest direction

After brainstorming, I compared the main options against what the capstone needs to do. I did not want to choose the most technically impressive option. I wanted to choose the option that best communicates the speculative tension of the project. This connected to the Double Diamond pattern of opening up possibilities and then narrowing toward a stronger direction. The Design Council describes the Double Diamond as a process of exploring an issue widely through divergent thinking, then taking focused action through convergent thinking (Design Council, n.d.).

From this comparison, the workplace simulation became the strongest capstone direction. The SaaS dashboard helps explain the system logic, but it still feels like a product demo. The workplace simulation makes the user perform inside the system, which makes the surveillance and AI authority easier to feel.

This changed my thinking. Ticker needs two layers:

The B2B AI senior explains the system. The workplace simulation communicates the feeling of the future.

Reverse brief activity: From SaaS product to workplace simulation

During the Week 8 reverse brief activity, I spoke with Nick about how this project should be shown for capstone. This conversation was important because it shifted my thinking away from only building a B2B SaaS tool. I had been focusing on how Ticker could plan projects, delegate tasks, generate SOPs, and connect that context to Focus Integrity. That made the system technically clearer, but it still risked being read as a normal productivity product.

The reverse brief activity helped me ask a different question:

How can I make the audience feel the future of Ticker, rather than only understand it as a software interface?

One idea that came from this conversation was to turn the capstone display into a workplace environment. Instead of placing Ticker on one table as a product demo, the whole space could be staged as a small office or work booth. The participant could sit at a dedicated computer, receive a task from the AI senior, and try to complete it while Ticker watches their task alignment.

A side television or large monitor could show the workplace dashboard. This could include the participant's Focus Integrity score, live task-alignment graph, warning state, and evidence summary. The dashboard could make the surveillance visible to both the participant and the audience. This changes the project from a private app interaction into a public workplace condition. The participant is no longer only using Ticker. They are being watched by Ticker.

The most provocative version of this idea was that if the participant does not stay aligned with the task, the system could escalate within five minutes. It could first show a soft warning, then a stronger warning, and finally a full-screen “YOU ARE FIRED” state. This would be a staged and fictional outcome, not a real judgement of the participant. The point is not to assess the participant's ability. The point is to make people feel how uncomfortable it would be if AI systems became responsible for assigning, monitoring, and evaluating human work.

This made the project feel much more capstone-aligned. The future I am showing is not just “AI helps people focus”. It is a future where AI becomes a workplace authority. It gives tasks, watches behaviour, produces scores, and turns effort into evidence.

By making the audience perform inside that system, the project can provoke a stronger question:

Would people accept AI surveillance if it made work more efficient, measurable, and easier to manage?

This also helped me understand what Ticker should become visually. The final outcome may need more than a dashboard. It may need a spatial setup: a work desk, a monitor, a public management screen, a live Focus Integrity graph, an AI senior voice or chat panel, and later a Ticker Cube object. The exhibition should make the user feel the tension between productivity support and surveillance.

However, this idea also raises ethical concerns. If the system uses a camera or shows a face feed, the participant needs clear consent and the experience should avoid storing or exposing real personal data unnecessarily. Employment New Zealand states that employers must comply with the Privacy Act 2020 and privacy principles when collecting, storing, using, and sharing employee-related information. It also notes that workplace monitoring may affect morale if employees feel less trusted (Employment New Zealand, 2025). This means the capstone experience should be staged carefully, with clear signage, visible consent, no hidden recording, and simulated or temporary evidence wherever possible.

The key learning from this activity was that Ticker should not be shown only as a product. It should be shown as an experience. The strongest capstone version is not just a screen that explains surveillance. It is a workplace simulation where the audience can feel how quickly support becomes control.

Experiment idea: Workplace simulation prototype

This experiment will test... whether Ticker becomes more understandable and impactful when it is experienced as a workplace simulation rather than viewed as a normal software demo.

It connects to my reverse brief because... the project is about AI becoming a work authority. A spatial workplace setup makes this authority more visible than a flat dashboard.

I will do this by... creating a small staged work environment with a dedicated computer, AI senior panel, public Focus Integrity dashboard, live graph, and a fictional “YOU ARE FIRED” fail state.

What I hope to learn is... whether a spatial and interactive installation communicates the speculative future more strongly than a normal product interface.

Reverse brief V2 after the discussion with Nick

This revised reverse brief helped me separate the system logic from the exhibition logic. The B2B AI senior system explains how Ticker works. The workplace simulation explains why the future matters.

Preparing the Crit 2 slide deck

The second major part of Week 8 was preparing the Crit 2 slide deck. This was important because the project had become more complex. Ticker was no longer only a Pomodoro timer, focus app, or social accountability system. It was becoming a speculative AI work system.

The slide deck helped me translate the Week 8 pivot into a critique-ready argument. Instead of only showing screens, I used the slides to explain why the direction changed, how the AI senior relationship works, and what ethical questions the project creates.

This made the slide deck part of the design process. I was not only preparing a presentation. I was creating a test object for critique. If people understood the slide deck, then the AI senior direction was becoming clearer. If they misunderstood it as only another productivity tool, then the project gap was still weak.

Reflection on Action

Designing Ticker as an experience

The most important shift this week was realising that Ticker should not only be developed as software. It needs to be designed as an experience. Before this, I was still thinking like a builder: if the system logic works, then the project is stronger. But the conversation with Nick helped me realise that capstone needs more than working logic. It needs a clear, memorable, and discussable future scenario.

The B2B AI senior model gave Focus Integrity richer task context, but the workplace simulation gave the project a stronger form. If the audience enters a staged workplace, receives a task, sees their Focus Integrity score displayed publicly, and experiences a fictional firing state, the project becomes much harder to dismiss as a normal productivity tool. It becomes a future condition that people can feel, question, and debate.

This also changed how I understood the role of speculation. I do not need to prove that Ticker is a product that should exist. I need to make the future visible enough that people can decide whether they want that future or not.

What became stronger

The Week 8 pivot made the project stronger because it gave Focus Integrity a clearer foundation. In Week 7, the system judged behaviour mainly from tick data. In Week 8, the system started to define the task before judging the behaviour.

This changed the role of Focus Integrity. It was no longer only an activity detector. It became a task-alignment estimate.

This made the concept more believable. A future where AI simply watches workers is easy to imagine, but also quite simple. A more interesting future is one where AI becomes useful enough that people accept its authority. The AI senior does not only watch. It organises, explains, assigns, checks, and supports. That is what makes it more powerful and more dangerous.

What became problematic

The same pivot also created a new problem. If Ticker can clarify briefs, generate plans, assign workers, write SOPs, and check behaviour, then it begins to move from support into control.

This tension is important for the project. Ticker should not be framed as simply good or bad. The most interesting part is that it can be helpful and invasive at the same time. It can make work clearer, but it can also reduce freedom. It can make Focus Integrity fairer, but it can also make surveillance easier to accept.

The Motion problem

Preparing the Crit 2 deck also made me realise that Ticker was moving close to existing AI productivity products. Motion already presents itself as an AI project-management platform that can automate project movement, prioritisation, capacity planning, progress visibility, and manager follow-up (Motion, n.d.). This means Ticker cannot simply be another AI planner.

This helped me sharpen the project gap. Ticker should not replace Motion, Jira, Slack, Calendar, or other work tools. Instead, Ticker can use those tools as context sources. Its role is not to become the whole workplace system. Its role is to create a task-aware verification layer that makes the future of AI-managed work visible.

Theory

Research grounding: task context, structured AI, and the market gap

The Week 8 pivot needs to be understood as more than a feature expansion. It is a shift from behavioural focus detection to task-context infrastructure. In Week 7, Ticker could collect activity signals such as app use, screen change, input activity, and camera presence, but those signals did not always explain whether the behaviour was meaningful in relation to the task. This is why Week 8 moved the project toward the idea of an Artificial Intelligence (AI) senior. The AI senior does not only watch the worker. It helps define the work before Focus Integrity judges whether behaviour aligns with it.

This shift is important because task alignment cannot be judged from raw activity alone. Chrome could be distraction, research, documentation, or testing. YouTube could be entertainment or a tutorial. A static screen could mean inactivity, reading, thinking, or waiting for code to compile. Without a clearer task contract, Focus Integrity risks becoming a generic activity score. With task context, it can become a more specific task-alignment estimate.

Structured AI output became a key research point this week. If the AI senior only produces conversational text, the output is hard to use inside the system. For Ticker, the AI needs to generate structured objects such as task title, expected output, expected tools, allowed behaviours, ambiguous behaviours, clarification triggers, and success criteria. OpenAI's structured output documentation explains that schema-based responses can make model output follow a defined structure rather than returning loose text (OpenAI, n.d.). Vercel's AI SDK documentation also notes that language models can produce incorrect or incomplete structured data, so schemas and validation are needed when generating structured objects (Vercel, n.d.). This directly changed how I understood the AI senior. It should not only “suggest a plan”. It should create a reviewable task contract that can later be used by Focus Integrity.

This also helped me clarify the market gap. Motion already presents itself as an AI project-management platform that automates project movement, prioritisation, capacity planning, deadline prediction, and manager visibility (Motion, n.d.). ActivTrak positions itself as a work-intelligence platform that captures behavioural activity such as hours worked, schedule adherence, location-policy compliance, and app or website usage, then analyses productivity trends and team performance (ActivTrak, n.d.). These examples show that there are already strong tools for AI planning and workplace analytics. Therefore, Ticker should not try to become another Motion or another ActivTrak. Its gap is task-aligned verification: using task context to decide whether a focus session appears aligned with what the worker committed to doing.

This market comparison made the Week 8 direction clearer. Ticker should sit between project-management tools and monitoring tools. It should use task data from planning systems, but its main role is to ask whether observable behaviour during a focus block matches the declared task context.

Human-in-the-loop as a design requirement

The AI senior also needs human oversight. If the system assigns workers, generates SOPs, or decides whether a user is off-task, it should not act silently. If AI clarifies briefs, generates plans, assigns workers, and creates Standard Operating Procedures (SOPs), then the system is making decisions that affect people. LangChain's human-in-the-loop documentation describes a model where agent actions can pause for human review, allowing the human to approve, edit, reject, or respond before the action continues (LangChain, n.d.). This is useful for Ticker because the AI senior should not silently become the boss. It should pause at moments where its decision affects a worker's task, score, evidence, or assignment.

Human oversight also cannot be symbolic. Sterz et al. (2024) argue that effective human oversight requires the human to have causal power, epistemic access, self-control, and fitting intentions. In Ticker terms, this means the manager or worker must not only see that “AI made a decision”. They need enough information and control to change that decision. A worker should be able to correct a task contract, challenge a Focus Integrity judgement, and explain ambiguous behaviour before the system turns it into a score.

This changed my Week 8 design requirement. Ticker's AI senior needs review points. The manager should approve or edit generated project plans. The worker should approve or clarify their focus contract. Focus Integrity should ask questions when evidence is uncertain. This keeps the system closer to “AI support” and further from “AI command”.

Workplace privacy and trust

The workplace version of Ticker needs to be grounded in privacy, especially because this project is situated in Aotearoa New Zealand. Employment New Zealand states that employers must comply with the Privacy Act 2020 and the Privacy Principles when collecting, storing, using, and sharing employee-related information (Employment New Zealand, 2025). It also warns that monitoring, recording, or filming employees can affect morale and productivity because employees may not feel trusted when they are monitored at work (Employment New Zealand, 2025). This directly applies to Ticker because Focus Integrity may involve app activity, browser activity, screen evidence, camera presence, and task records.

The Office of the Privacy Commissioner explains that the Privacy Act 2020 gives New Zealanders rights around knowing when information is collected, having it used and shared appropriately, having it kept safe, and accessing their information (Office of the Privacy Commissioner, n.d.). This means Ticker cannot treat evidence collection as a hidden technical layer. If the system collects task, screen, app, or presence data, users need to know what is collected, why it is collected, who can see it, how long it is kept, and how they can correct wrong information.

This research made Week 8 more ethically grounded. The problem is not only whether Ticker can judge work. The problem is whether it can do so in a way that is visible, limited, explainable, and contestable.

Speculative design: when AI becomes useful enough to become authority

The Week 8 pivot also strengthened the speculative value of the project. Dunne and Raby (2013) argue that speculative design is not only about predicting the future, but about using possible futures to question present assumptions. In this case, Ticker is not only asking whether AI can improve productivity. It is asking what kind of work culture emerges when AI becomes useful enough to plan, assign, guide, question, and verify human effort.

This is more interesting than a simple “AI surveillance is bad” argument. The AI senior is ethically uncomfortable because it could be genuinely helpful. It can clarify vague work, reduce confusion, create fairer expectations, and help users stay aligned with their own goals. However, those same benefits can make AI authority feel normal. If the AI defines the task, writes the SOP, assigns the worker, and checks the worker, then it starts to shape what counts as valid work.

This became the core Week 8 theoretical insight:

Ticker is not scary because it is obviously dystopian. It is scary because it might feel helpful enough that people accept AI authority over their work behaviour.

Preparation

Updated Crit 2 plan

For Crit 2, I need to test both the system logic and the capstone format.

This will help me decide whether Ticker should remain mainly as a software demo or become a staged workplace experience where the audience feels the pressure of AI-managed work.

What I will test next

The next step is to bring the AI senior direction into Crit 2 and use the feedback to decide what the strongest next prototype should be.

Conclusion

Week 8 was important because it changed the project from Focus Integrity as detection into Focus Integrity as task-aware verification. Week 7 showed that behavioural data alone was not enough. This week, I responded by designing the AI senior workflow: a system that clarifies briefs, generates plans, assigns work, creates SOPs, and turns tasks into focus contracts.

This made the project stronger because Focus Integrity gained context. Instead of only watching apps, typing, screen changes, and camera presence, Ticker could begin by understanding what the user was meant to do. This made the judgement feel more meaningful.

However, the pivot also made the project more ethically complex. If AI defines the work, assigns the person, writes the process, and checks whether the worker follows it, then AI becomes more than a tool. It becomes a work authority.

Preparing the Crit 2 slide deck helped me see this clearly. The deck was not only a presentation. It was a way to test the project's argument. I needed to know whether people would understand Ticker as a task-aware AI senior, whether the system felt different from Motion, and whether the ethical risk of AI-managed work was strong enough.

The research also helped me understand why Week 8 was not only a technical pivot. Structured AI output made the workflow more buildable. Human-in-the-loop AI made the system more responsible. Workplace privacy research made the surveillance risk harder to ignore. Motion and ActivTrak gave me a sharper competitor boundary. Speculative design helped me frame the project as a question about future AI authority, not only as a productivity product.

By the end of Week 8, the project had a clearer direction but also a sharper tension. Ticker should not become another planner, another to-do app, or another workplace monitoring tool. It should become a speculative system that asks what happens when AI becomes useful enough to organise, guide, question, and verify human effort.

Week 8 shifted Ticker in two ways. Technically, it moved from Focus Integrity detection into AI task context and delegation. Spatially, it moved from a software product demo into a possible capstone workplace simulation.

References

• ActivTrak. (n.d.). Work intelligence for productivity optimization. Retrieved May 28, 2026, from https://www.activtrak.com/
• Baldwin-Ramult, L. (2026). DES304: Emerging technologies stream brief[Course brief]. University of Auckland.
• Design Council. (n.d.). Framework for innovation. https://www.designcouncil.org.uk/resources/framework-for-innovation/
• Dunne, A., & Raby, F. (2013). Speculative everything: Design, fiction, and social dreaming. MIT Press.
• Employment New Zealand. (2025, August 14). Employee privacy. New Zealand Government. https://www.employment.govt.nz/fair-work-practices/employee-privacy
• LangChain. (n.d.). Human-in-the-loop. Retrieved May 28, 2026, from https://docs.langchain.com/oss/python/langchain/human-in-the-loop
• Motion. (n.d.). Motion's AI project manager. Retrieved May 28, 2026, from https://www.usemotion.com/features/ai-project-manager
• Office of the Privacy Commissioner. (n.d.). Privacy Act 2020. Retrieved May 28, 2026, from https://www.privacy.org.nz/privacy-principles/
• OpenAI. (n.d.). Structured model outputs. Retrieved May 28, 2026, from https://platform.openai.com/docs/guides/structured-outputs
• Sterz, S., Baum, K., Biewer, S., Hermanns, H., Lauber-Rönsberg, A., Meinel, P., & Langer, M. (2024). On the quest for effectiveness in human oversight: Interdisciplinary perspectives. arXiv. https://arxiv.org/abs/2404.04059
• University of Edinburgh. (2024). The Integrated Reflective Cycle. Reflection Toolkit. https://reflection.ed.ac.uk/reflectors-toolkit/reflecting-on-experience/the-integrated-reflective-cycle
• Vercel. (n.d.). Generating structured data. AI SDK. Retrieved May 28, 2026, from https://ai-sdk.dev/docs/ai-sdk-core/generating-structured-data