<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by NTX Studio on Medium]]></title>
        <description><![CDATA[Stories by NTX Studio on Medium]]></description>
        <link>https://medium.com/@ntx.studio?source=rss-7de3e5088dff------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*kc4-5zdl_x6LCV3uFWSncQ.jpeg</url>
            <title>Stories by NTX Studio on Medium</title>
            <link>https://medium.com/@ntx.studio?source=rss-7de3e5088dff------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 12:57:41 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ntx.studio/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[How to work with AI]]></title>
            <link>https://medium.ntx.studio/how-to-work-with-ai-e61b596f8218?source=rss-7de3e5088dff------2</link>
            <guid isPermaLink="false">https://medium.com/p/e61b596f8218</guid>
            <category><![CDATA[future-of-work]]></category>
            <category><![CDATA[human-ai-collaboration]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[organizational-learning]]></category>
            <dc:creator><![CDATA[NTX Studio]]></dc:creator>
            <pubDate>Fri, 13 Nov 2026 12:21:43 GMT</pubDate>
            <atom:updated>2026-11-13T12:21:43.646Z</atom:updated>
            <content:encoded><![CDATA[<p>Consider a woman approaching retirement after decades of factory work. She has paid her contributions, kept her documents, and contacted the pension provider to find out what happens next. When she calls again, the employee answering cannot see what the previous colleague told her. She starts from the beginning. For the institution, this is another enquiry. For her, it concerns how she will live when her wages stop.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Ms20NJFlUMW1Qwm56HLKHA.jpeg" /></figure><p>The customer-service manager reviewing the case finds that everyone followed the procedure. Each employee supplied the required information and closed the contact. Nobody established whether the woman could now proceed with her claim. The manager wants to improve the service, and AI offers some promising help: recovering case histories, finding relevant guidance, and reducing the work of moving information between systems. What should the team be able to do with that help?</p><p>Our <a href="https://medium.ntx.studio/what-should-ai-help-us-become-89676dc3ad8a">first essay</a> distinguished an obstacle from an opportunity to practise. A designer may want help learning to choose colours. The woman claiming her pension has little reason to master the provider’s filing system. For her, repeating her occupational history is an obstacle the institution should remove. For the employee, understanding a difficult claim may be practice that develops their judgement. The manager needs to recognise which effort helps whom and arrange assistance accordingly.</p><p>The service should give people time to ask questions and understand their options, including when their work history is incomplete or a claim has to be refused. Managers need to make room for that attention and help employees learn to handle difficult conversations.</p><h3>Remembering the person</h3><p>With her contributions and previous enquiries available to the authorised employee, the conversation can begin with what still needs attention. The provider has already received much of the information it needs. Connecting those records lets staff build on earlier work and gives the participant continuity between conversations. The manager can then examine where the case stalled and which part of the service needs to respond.</p><p>Retirement brings questions about income and paperwork, alongside the end of a familiar working life. An employee who understands the circumstances can explain what will happen and check what remains uncertain. The participant has someone to ask when she should expect payment and what to do if it does not arrive. A personalised greeting loses much of its charm if the next question is one the customer has answered twice already.</p><p>The manager also needs to know what reassurance means to different participants. One may want help planning the next stage of life; another simply needs to know when the money will arrive. Staff can ask, listen, and adjust their response. The employment history gives them a useful starting point, and the conversation helps them understand what matters to this person now. Access to records becomes valuable through the attention it allows the employee to give.</p><p>AI could bring relevant information together and help staff interpret unfamiliar requirements. Participants could describe their situation in their own words. An employee could ask for an explanation of an unusual provision, open the underlying guidance, and check how it applies. Case summaries would carry links to the records, with access set according to each employee’s role.</p><p>The manager can learn from how employees already handle these conversations. One colleague may have found a clearer way to explain a requirement; another may recognise why apparently straightforward applications keep returning. AI can help compare cases and prepare alternative explanations. Sharing what works gives less experienced staff something to learn from, while recognising the practical knowledge their colleagues have developed.</p><p>The team can also draw on knowledge beyond its own workplace. <a href="https://birdsnfrogs.github.io/2026/09/12/Felix_qui_potuit_rerum_cognoscere_causas.html">AI can make accumulated cultural knowledge more accessible</a>, bringing established approaches into reach of people encountering a problem for the first time. An agent might recover an unfamiliar method, connect it with the current difficulty, or suggest an alternative worth trying. The manager and staff can explore those possibilities together, drawing on their understanding of the participants, the regulations, and the resources available.</p><p>Suppose the team initially proposes better summaries for transferring enquiries. Testing the idea with staff reveals that many transfers happen because the first employee lacks access to information another department holds. A clearer summary would help, but giving the first employee appropriate access might remove the transfer altogether. Investigation changes what the manager thinks the service needs. AI can contribute to that discovery, as well as help deliver the eventual improvement.</p><p>When asking AI to improve the workflow, the manager should keep the proposed targets open to discussion. An agent told to reduce transfers may recommend keeping every case with the first employee, even when specialist advice is needed. Another asked to critique that proposal may accept the same target. Staff need a way to explain why a particular transfer serves the participant and which conditions make it necessary. A technically competent answer can still advance a poor brief.</p><h3>When the procedure fails</h3><p>Procedures give employees a reliable starting point. They preserve knowledge, support consistency, and protect participants from arbitrary decisions. A beginner can draw on decisions experienced colleagues have already worked through. Authority is often divided along the same lines: specialists establish the procedure, and frontline staff apply it. When a case falls outside the expected categories, the employee handling the call needs access to someone who can help reconsider it.</p><p>A longer conversation can then count against performance even when it prevents another call. Asking a different department to examine a case creates work outside the prescribed route. The safest response for the employee is to supply the approved answer and move on. Advising staff to think critically does little when the company disciplines them for taking the time to do it.</p><p>Customer service is already changing under pressure from automation. An <a href="https://www.reveliolabs.com/news/ai-and-work/after-a-decade-of-growth-global-call-center-employment-is-shrinking">employment analysis of call centres</a> reports global headcount 5.3% below its December 2023 peak. Across twelve middle-income countries, the contraction largely reflects slower hiring. The report considers AI adoption alongside post-pandemic adjustment and wider cost-cutting. For people entering office work, reduced recruitment can close a route through which they would have gained experience and developed their skills.</p><p>The same analysis finds that only 10.8% of observed movers from call centres in those countries reached higher-paying technical support, customer success, or software and data roles within twelve months. Moving into more skilled work takes preparation and an opportunity to use it. Employers proposing that transition need to provide training, supervised practice, and a credible route into the new role, with support for workers while they learn.</p><p>Faced with a growing queue, the manager may defer training and investigation to get more calls answered today. The unresolved causes generate further calls, leaving even less time to address them. This is the capability trap described in the first essay, now operating inside the service. AI can increase the number of contacts handled while leaving that cycle intact. Breaking it requires using some of the saving to reduce recurring problems and develop the team’s ability to resolve them.</p><p>Some Chinese government offices have established “办不成事”, or “can’t get things done”, complaints windows for matters ordinary service channels have failed to resolve. Their <a href="https://www.wxlx.gov.cn/doc/2023/04/23/3939233.shtml">operating rules</a> provide for knowledgeable staff, coordination across departments, and escalation when responsibility is unclear or the problem is difficult. Recurring failures are to be examined so that the service itself can improve.</p><p>The staff member receiving the complaint can bring the relevant departments into the case, establish who is responsible, and agree a time for a response. The participant then has someone to contact who can explain the decision or pursue the next step. Recording the outcome also gives the institution a way to identify recurring failures and decide which procedures need changing.</p><p>In <a href="https://www.zs.gov.cn/zstzz/gkmlpt/content/2/2604/post_2604516.html">one reported case</a>, older residents without local household registration had to travel to a city centre to obtain senior bus cards. Staff worked with the card company to make applications available at the town’s service centre and sixteen village or community locations. Following the complaint led them to reconsider where the service was offered. The resulting change addressed the difficulty for future applicants as well as the people who had raised it.</p><p>Our pension-service manager can give staff a similar route for pursuing discrepancies in employment records. AI may help find related cases or assemble supporting information. The employee needs a named contact in the relevant department and a response they can explain to the participant. When the same discrepancy recurs, the manager can use those cases to investigate its cause, agree a correction, and give the team revised guidance.</p><h3>Learning to serve</h3><p>Giving staff greater responsibility raises a question about how they acquire the understanding to use it. A beginner can ask AI to explain a rule, explore how it applies to the case, and identify what remains to be established. Working through the explanation with an experienced colleague helps them learn which facts deserve attention. Assistance becomes part of the practice through which an employee develops the understanding needed to question later recommendations.</p><p>The <a href="https://www.theguardian.com/commentisfree/2026/jul/24/should-you-use-ai">distinction between doing work and going to the gym</a> asks whether an activity is intended to produce a result or develop the person doing it. A job often serves both purposes. Resolving the woman’s claim matters now. Understanding why a particular document establishes entitlement can help an employee resolve other claims later. The manager has a responsibility for both, even if the monthly report counts only completed cases.</p><p>Research gives the manager reason to examine how that practice changes with AI. A <a href="https://arxiv.org/abs/2506.08872">small essay-writing study</a> reported poorer immediate quotation recall and lower perceived ownership among AI-assisted writers. A <a href="https://www.mdpi.com/2075-4698/15/1/6">survey of 666 people</a> associated greater AI use and cognitive offloading, or delegating mental tasks to tools, with lower self-reported critical-thinking scores. A <a href="https://doi.org/10.1016/j.chbr.2026.101102">cross-domain review</a> identifies a broader pattern of short-term offloading and uncritical reliance. The duration and extent of changes in skill remain open research questions; the immediate effects already warrant attention when designing work.</p><p>A <a href="https://doi.org/10.1073/pnas.2422633122">randomised school mathematics study</a> shows how the design of assistance can affect learning. Ordinary chatbot assistance improved practice results but reduced subsequent unaided exam performance by 17% relative to the control group. Learning-oriented safeguards largely removed that deficit. In this classroom setting, the way students received help affected what they could subsequently do for themselves. The manager faces a related design question when deciding how staff should use assistance during cases they need to learn to handle.</p><p>After a case, the manager can ask the employee to explain which evidence mattered and why. Comparing two similar applications with an agent and a senior colleague can reveal a distinction they initially missed. Beginning to recognise those distinctions for yourself brings satisfaction, alongside the benefit to the next person who needs help.</p><p>This takes better feedback than “use your judgement”. A useful correction identifies the missing consideration and explains what it could change. We take considerable care to give machines detailed errors and another attempt. Junior colleagues are sometimes expected to make do with “you should have known”. The manager can serve the team by making the reasoning available and allowing another attempt, with support appropriate to the consequences of getting it wrong.</p><p>Learning also depends on what colleagues continue to practise. A <a href="https://arxiv.org/abs/2609.03344v1">theoretical model of human–AI coupling</a> uses illustrative assumptions about unaided competence to explore how institutions and peers sustain independent reasoning. Within the model, weakening those supports can produce dependence that persists after adoption pressure eases. Extensive AI use can also coexist with autonomy when the conditions support it. For the manager, preserving opportunities to explain and examine decisions helps sustain the expertise future employees will learn from.</p><p>The manager can arrange opportunities to examine a difficult case, receive an explanation, and try applying what has been learnt. AI can participate throughout, helping an employee explore an unfamiliar rule or compare possible decisions. An experienced colleague can judge when closer supervision is needed and when the employee is ready to take more responsibility. Support develops with the person, keeping the work useful to the participant while making room for the employee to grow.</p><h3>What the manager owes the team</h3><p>The manager’s own judgement will sometimes be wrong. Experience can make a familiar solution look more convincing than it deserves. Staff need to be able to question the proposed approach while their findings can still affect it. If an early trial reveals that employees need access to another department’s records, the project must have time and authority to address that dependency before the new service is launched.</p><p>An early trial could examine whether employees can recover the case history, identify missing evidence, and explain the next step to a participant. A later trial could test whether the complete service is dependable. AI can help prepare alternatives and examine what each trial would establish. Each round should address an uncertainty that matters to the next decision, with a stopping point that allows the team to act on what it has learnt.</p><p>Operational standards should reflect what the team is trying to achieve. Handling time matters, especially for simple enquiries. So do repeat contacts, errors, and unresolved cases. Fewer complaints may indicate improvement or a channel people have given up using. The team needs to examine what its measures mean, including whether an apparently successful AI recommendation has made the underlying problem harder to see.</p><p>Useful discoveries should become part of the organisation’s memory. Recording why a transfer proved unnecessary gives another employee something to recognise in a different case. The explanation can inform an agent, support training, and give the manager grounds to revise a procedure. What the service learns should remain available after the employee who noticed it has moved on.</p><p>The manager has to make room in the schedule for an employee to investigate a recurring problem, and arrange cover for the calls that will keep arriving. The employee may also need help from an experienced colleague or someone who maintains the records. Budgeting for those hours is part of deciding whether the proposed service improvement is worth pursuing.</p><p>In a business, shareholders have a stake in both immediate returns and the service’s continuing viability. Decisions about staffing and investment will involve real disagreements over who benefits and when. We should make those choices visible. Time released by automation becomes an opportunity to develop people only if they are allowed to use it that way. Serving employees is part of the organisation’s purpose, as well as a condition of their serving others well.</p><p>The woman arranging her retirement should be able to leave knowing what happens next, without preparing to repeat the whole conversation. The employee helping her should understand enough to explain the decision and have somewhere to turn when the records do not fit. Helping both of them is the manager’s work. AI should help the team resolve the claim and improve the service. The institution must give them the time and authority to use that help well.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e61b596f8218" width="1" height="1" alt=""><hr><p><a href="https://medium.ntx.studio/how-to-work-with-ai-e61b596f8218">How to work with AI</a> was originally published in <a href="https://medium.ntx.studio/">Studio Of Thought</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>