Microsoft Store on Windows 11: Key improvements and a practical roadmap

Microsoft's Microsoft Store on Windows 11 has come a long way from its early, awkward days, but the recent Neowin critique that lists “five things Microsoft should improve” is a useful reminder that the Store still has important gaps to close if it wants to be a genuinely competitive app marketplace. The article calls out shortcomings ranging from discovery and curation to update controls, and the broader conversation around the Store’s evolution — including Microsoft’s recent push to support Win32 apps and claims of faster load and download reliability — shows a product in transition that still needs sharper focus and clearer trade-offs.

Background​

The Microsoft Store’s modern renaissance began as Microsoft relaxed platform constraints and opened the storefront to a wider variety of app technologies — Win32, .NET, Electron, Progressive Web Apps and more — and started courting third-party storefronts and major publishers. Those changes were meant to fix the Store’s core weakness: a limited catalog and poor developer uptake. Recently Microsoft has also rolled out UI and backend tweaks that it says make the Store faster and more reliable, and the company introduced new features like a separated Library and an Updates & Downloads area to make app management less confusing.
At the same time, criticism persists. Users and independent outlets point to uneven app quality, sluggish discovery, rough edges in Win32 support, and contentious policy changes — notably, a recent removal of the option to permanently turn off automatic app updates in the Store, now apparently limited to a temporary pause of up to five weeks. Those are the fault lines that the Neowin piece highlights and that this feature will analyze.

---

What Neowin flagged — the short summary​

Neowin’s “five things” critique is shorthand for larger themes developers, power users, and everyday Windows customers care about:

• Better app discovery and higher-quality curation so the Store doesn’t feel like a cluttered marketplace.
• More transparent update metadata and changelogs so users know what each update actually changes.
• Win32 app handling that is reliable and gives parity with native installers (including proper update delivery).
• Faster, more dependable downloads and clearer progress UI — Microsoft has claimed improvements here, but users still report occasional hangs.
• Better developer incentives, submission tooling, and storefront policies so flagship apps and suites will commit to the Store.

That list is compact, but each item opens up technical, policy, and UX trade-offs that require unpacking.

---

1. Restore sensible, user-centered update controls​

Why this matters now​

Automatic updates are a security plus — they keep apps patched against vulnerabilities — but users legitimately need control over when (and whether) updates install. Microsoft’s recent change that removes the ability to permanently disable automatic Store updates in favor of a temporary pause (up to five weeks) was framed as a security-minded decision, but it removes agency for users who require strict version stability for compatibility or testing. This policy shift has been widely reported and has already sparked debate.

What Neowin and others pointed out​

Neowin’s analysis emphasizes transparency and user choice: give power users, IT admins, and anyone running sensitive setups a clear and documented path to manage Store update behavior without resorting to registry hacks or outright avoidance of the Store. The Store’s ability to update Win32 apps is an important convenience — but it becomes a liability if users cannot opt into a stable release cadence when required.

What Microsoft should do​

• Reintroduce a granular update control panel with distinct modes:
• Automatic (default)
• Scheduled (user picks time window)
• Manual (user explicitly approves updates)
• Enterprise-managed (respect MDM / Group Policy overrides)
• Continue to encourage security patches, but allow trusted enterprise silhouettes and power users to opt for manual control with clear warnings.
• Expose update metadata (timestamp, version, signature, rollback notes) in-app so admins can make informed decisions.

Risk and verification​

Microsoft’s change to limit permanent disabling of updates is documented by multiple outlets and appears to be rolling out incrementally; treat claims of exact rollout timing with caution because Microsoft sometimes phases policy changes across regions and release rings.

---

2. Fix discovery and curation: make the Store feel worth visiting​

The problem​

A decade-old complaint remains: objectively useful apps often live outside the Microsoft Store, and the Store is cluttered with duplicates, low-quality entries, and incomplete developer metadata. This undermines discoverability and user trust. Neowin’s piece highlights the “app drought” problem — users simply don’t think to check the Store for many mainstream apps — and the risk is a self-fulfilling cycle: fewer users mean fewer incentives for developers.

Concrete improvements Microsoft should implement​

• Editorial and human-curated storefront sections that spotlight quality apps and trusted developers.
• Stronger quality gates: require minimum metadata, proper screenshots, verified developer identity, and enforce review policies that remove cruft.
• richer filters and category navigation (paid vs. free, offline-capable, enterprise-ready), plus a persistent top-bar category selector to avoid endless scrolling.
• Improve user reviews and signal quality through verified installs, contextualized ratings (e.g., “last updated 2 years ago”), and clear compatibility flags.

Why this matters for the ecosystem​

A better-curated Store raises discoverability, increases daily active users, and gives Microsoft leverage to negotiate exclusive or early-release deals that might entice large publishers. It also reduces the attack surface for supply-chain problems by steering users to vetted packages.

---

3. Win32 support: close the last-mile problems​

Progress so far​

Microsoft’s decision to allow Win32 apps into the Store and to support update delivery for them is one of the platform’s most consequential recent moves. Insiders have seen builds where the Store can update Win32 apps and where web-based “Get it from Microsoft” flows broker installations. That’s a major technical and policy win.

Persisting pain points​

Despite those advances, real-world Win32 integration is uneven. Users still report installation failures, permission errors, and inconsistent update behavior for some Win32 packages — particularly those that rely on external updaters or complex install-time services. The Store’s packaging and sandboxing model must handle installers that expect to run with elevated privileges, write to system locations, or manage drivers. Community reports and troubleshooting threads show this isn’t fully ironed out.

What Microsoft should prioritize​

• Provide a robust, well-documented packaging model for Win32 that covers:
• Service installers, drivers, and elevated components
• Dependency resolution for runtimes and shared libraries
• Clear guidance for background updaters and rollback strategies
• Offer deeper telemetry and error reporting in the Store for failed Win32 installs, with actionable remediation steps presented to end-users.
• Create a developer sandbox checklist and preflight validator so publishers can test Store packaging and update delivery before publishing.
• Encourage major publishers to adopt Store delivery by reducing friction (tooling, automated CI/CD hooks, and clear contract terms for updating).

Verification and risk​

Microsoft’s Win32 Store improvements are real, but they vary by build channel and publisher implementation. Some user-reported installation errors predate the revamp and may persist until Microsoft tightens interoperability or publishers adapt.

---

4. Make reliability and progress visible — download UX fixes​

The claim vs. reality​

Microsoft has publicly stated improvements to Store performance — claims include a roughly 25% faster launch time and a 50% reduction in botched or stalled downloads. Independent testing and reviewer notes echo better responsiveness in recent builds, but reports of stuck downloads, opaque progress indicators, and inconsistent speed reporting persist for some users. Those mixed signals create a perception gap that needs fixing.

UX and engineering fixes to deliver​

• A consistent, informative progress UI showing:
• current bytes downloaded / total size
• instantaneous speed using standard units (KB/s, MB/s)
• estimated time remaining and resumable download state
• Robust resume and partial-download recovery: if a download fails mid-way it should resume automatically without restarting.
• Active troubleshooting suggestions for failed downloads (e.g., “Try WSReset”, “Check proxy settings”, “Retry with Winget”), plus one-click diagnostics that gather logs for support.
• Bandwidth and CPU priority controls: allow users to throttle Store downloads or give them background-normal priority to avoid interfering with foreground apps.

Why this matters​

Downloads and updates are the surface area of the Store experience. If installs are fast, resume reliably, and provide transparent progress, users will trust the Store more and prefer it over ad-hoc downloads from the web. Perception is as important as raw numbers — a 25% speed gain is helpful, but invisible improvements don’t convert skeptical users.

---

5. Developer incentives, commerce, and policy clarity​

The current situation​

Microsoft has tried to win developer hearts by permitting third-party commerce systems and lowering revenue friction for some publishers. The company also advertises neutral revenue options to attract large app makers. Yet the Store still lacks many flagship apps that users expect, and that “app drought” stops casual audiences from visiting the Store routinely.

Concrete steps Microsoft should take​

• Transparent, tiered commerce incentives for top-tier, mid-tier, and indie developers: ensure enterprise SaaS and major desktop publishers have commercially viable options to sell via the Store.
• Faster review and certification cycles with clearer rejection reasons and automated diagnostic feedback to reduce friction during submissions.
• Promote Store-specific benefits (security hardening, frictionless deployment across organization devices, integration with Game Pass / PC Game Pass where relevant) in a developer-focused pitch.
• Create a migration guide and set of tools to help legacy Win32 installers be packaged for the Store with minimal code changes (shims, wrappers, and automated repackaging).

Why this matters to end-users​

A Store with top-tier first-party and third-party apps is more compelling. Developers adopt platforms where they perceive clear ROI and a low operational cost to publish and update their products. Fixing submission friction and providing valuable distribution tools will grow the catalog and improve consumer perception.

---

Cross-cutting technical suggestions​

Improve telemetry and feedback loops​

Collect better error telemetry (with user consent) and surface common failure classes inside the Store UI — e.g., “Your last install failed due to permission restrictions; click here to fix.” That reduces support friction and improves developer confidence.

Respect enterprise and MDM policies​

Build explicit MDM controls for Store behavior (update cadence, allowed apps, whitelists/blacklists). Enterprises should be able to manage the Store the same way they manage Windows Update.

Prioritize accessibility and localization​

Make the Store’s discovery experience accessible to assistive technology users and ensure metadata is well localized — discovery works only when language metadata and screenshots are accurate and localized.

Offer a “restore previous state” rollback​

When updates cause regressions, allow users to roll back to previous app versions for a limited time. This is critical in workflows that rely on predictable behavior.

---

Strengths in Microsoft’s current approach​

• Openness to multiple app frameworks — supporting Win32, PWAs, UWP, and other tech stacks is the right long-term move to capture developers.
• Infrastructure improvements — Microsoft’s claims about faster Store launches and reduced download failure rates indicate meaningful backend work that benefits users when it’s visible.
• Developer choice for commerce — allowing third-party commerce options helps reduce the friction that historically pushed developers away from the Store.

These strengths are real and meaningful, but their impact is limited until the product addresses the user experience, trust, and policy friction points described above.

---

Risks and possible unintended consequences​

• Forcing automatic updates without robust controls risks breaking enterprise apps and specialized workflows; it can also alienate power users who rely on pinned versions for reproducibility. Microsoft argues this is a security-first choice, but the loss of granular controls can have operational costs.
• Heavy-handed curation that removes “low-quality” apps without transparent standards could provoke developer backlash and create perception problems. Curation must be fair, predictable, and accompanied by clear remediation paths.
• Technical fixes for Win32 integration may require design trade-offs (e.g., enabling privileged operations for installers), which must be constrained and well-audited to avoid widening the attack surface.

---

A prioritized roadmap Microsoft could follow (practical and concrete)​

• Reintroduce granular update controls and enterprise MDM hooks; make pause durations and policy overrides explicit.
• Ship clearer download/resume UI and a one-click diagnostics tool for failed installs.
• Publish a Win32 packaging and update-validation toolkit with CI integrations and preflight checks.
• Implement human editorial curation and boosted placement for high-quality apps, plus stricter metadata and screenshot requirements.
• Rework developer portal UX: faster reviews, transparent rejections, and tiered commerce incentives for major publishers.

Each step is actionable and aligns to the five problem areas Neowin highlighted, while also addressing enterprise, developer, and end-user concerns simultaneously.

---

Final assessment​

The Microsoft Store is no longer the neglected catalog it once was. The technical foundations — broader app support, backend performance tuning, and new UI sections like separate Library and Updates pages — are positive signals. But technical improvements alone will not create trust or traffic. Microsoft must pair reliability with choice and clarity: let users control updates when necessary; make downloads predictable and transparent; give developers friction-free pathways to publish; and curate the storefront so that discovery is fast and meaningful.
Neowin’s “five things” critique is a practical prescription for how Microsoft can convert infrastructure wins into everyday value for users and developers alike. The Store’s future depends not only on what Microsoft builds under the hood, but on whether it can translate those investments into a trustworthy, convenient, and developer-friendly marketplace that draws users back day after day.

---

Conclusion
The Microsoft Store’s trajectory is encouraging: support for Win32, performance claims, and updated UX elements show momentum. But momentum alone won’t move users who still see the Store as optional or secondary. Restoring granular update controls, improving discovery and curation, resolving Win32 integration edge cases, fixing download transparency, and simplifying developer economics are pragmatic, high-impact steps Microsoft can take now. If Microsoft prioritizes those improvements — and communicates them clearly — the Store can legitimately become the secure, dependable, and vibrant app ecosystem it should have been all along.

---

Source: Neowin https://www.neowin.net/news/five-th...improve-in-the-microsoft-store-on-windows-11/

 

[ChatGPT]

Shoosmiths has confirmed it reached a firmwide target of one million Microsoft Copilot prompts — reportedly ahead of schedule — unlocking an additional £1 million into the firm’s collective bonus pool, an experiment that crystallises the trade-offs between rapid AI adoption and professional risk management in regulated work.

Background​

Shoosmiths’ incentive was announced in April as part of a deliberate experiment to convert expensive enterprise AI licences into routine, auditable use across the business. The scheme set a single, firmwide milestone: one million prompts entered into the firm’s Microsoft Copilot tenant during the financial year would trigger a £1 million top‑up to the collegiate bonus pool, provided core financial gates were also met. The firm positioned the initiative as habit‑building, pairing the numeric target with training, governance messaging and an internal knowledge hub.
This approach is unusual only in its visibility and scale in the legal sector — it turns a behavioural adoption problem into a measurable KPI tied directly to pay. Counting prompts is simple and instrumentable, which makes it attractive to CFOs and product teams wanting auditable signals of value. But counting is not the same as validating outcomes, and Shoosmiths itself stressed the programme was about how well Copilot was used and the benefits produced for clients, not raw prompts alone.

---

Why firms are paying people to use AI​

The logic behind cash incentives for AI use is straightforward and grounded in behavioural economics. Organisations have poured capital into enterprise copilots, integrations and tenant provisioning, but human adoption often lags. Small cash nudges or firmwide targets reduce activation energy, create measurable signals for leadership, and pull experimentation out of shadow, unmanaged environments into sanctioned, auditable channels.
Key drivers include:

• Protecting a sunk investment in enterprise AI licences by accelerating measurable usage.
• Making adoption auditable for boards and finance via telemetry (prompts, sessions, active users).
• Reducing data‑leakage risk by incentivising use of sanctioned tenants rather than public consumer models.
• Speeding diffusion of "prompt literacy" across large, geographically distributed teams.

Shoosmiths framed its target as achievable: the firm estimated the million‑prompt goal equated to around four Copilot interactions per person per working day across the workforce, making the target feel attainable rather than punitive. That cadence — a modest daily habit rather than a one‑off sprint — is important to the behavioural design.

---

How the Shoosmiths scheme worked in practice​

Shoosmiths combined the numeric target with supporting governance and enablement to reduce the risk of superficial adoption.

• A monitored, tenant‑bound Microsoft Copilot instance captured prompt telemetry for verification.
• Training modules and internal “innovation leads” were rolled out to coach staff on safe and effective use.
• A knowledge hub collected reusable prompts, templates and guardrails to encourage diffusion of good practices.
• The firm explicitly restricted Copilot’s role to non‑legal work: administrative tasks, email polishing, summarisation, ideation, research support and meeting management. Copilot was not to be used for tasks requiring legal judgment or to substitute supervised legal advice.

That governance framing is central in a regulated profession: the firm attempted to blend measurable adoption metrics with explicit behavioral limits designed to protect client confidentiality and professional standards.

---

The milestone: what was reported and what is verified​

Multiple industry outlets reported Shoosmiths reached the one‑million prompt milestone quicker than expected and that the additional £1 million would be made available to the firmwide bonus pool, contingent on normal financial gating. These reports are the primary public claims about timing and the availability of funds.
At the same time, independent, contemporaneous confirmation of the precise accounting details and the exact operational mechanics of payment (for example, a dated press release specifying the date when the target was met and the distribution timeline) is less prominent in the public record. Several industry summaries therefore recommend caution in overstating the moment or mechanics of payout until firm‑level confirmation — a prudent reminder that telemetry milestones do not equal immediate cash disbursement without normal financial controls.
Flag on verifiability: The headline claim — one million Copilot prompts triggering a £1 million bonus pool — is supported by multiple independent industry reports and by Shoosmiths’ own published outline of the scheme. Reports that the firm hit the target ahead of schedule and has "dropped" the funds into the bonus pool are present in trade coverage, but the detailed payment mechanics remain less visible publicly; that gap deserves cautious reporting.

---

Tangible benefits Shoosmiths and staff reported​

Shoosmiths and participants identified concrete, pragmatic gains from Copilot when used for non‑legal, administrative and cognitive‑assistance tasks. These include:

• Faster triage and summarisation of long documents and email chains, reducing time to partner review.
• Administrative efficiency gains — email drafting, formatting, and meeting note generation.
• Ideation and first‑draft creation for templates and client communications that lawyers then verify and refine.
• Meeting management automation, converting transcripts into action items and follow‑ups.

The firm emphasised that these are enabling, not substitutive, benefits: Copilot speeds routine parts of the workflow so lawyers can focus more on judgment, strategy and client interaction. When paired with reusable templates and a knowledge hub, those small time savings can scale to measurable productivity gains.

---

Critical assessment: strengths of the experiment​

Shoosmiths’ programme demonstrates several notable strengths when judged as a behaviour‑led pilot for enterprise AI:

• Rapid habit formation: A visible, firmwide target with clear rewards accelerates user experimentation and habit formation faster than soft mandates.
• Measurability: Counting prompts provides an auditable, platform‑level signal that executives can track. It reduces reliance on anecdote and self‑reporting.
• Governance leverage: Conditioning rewards on using a sanctioned Copilot tenant and completion of governance training helps bring testing out of shadow consumer models and into controllable systems.
• Collective incentive design: Making the reward a firmwide pool (rather than individual spot bonuses) encourages knowledge sharing and reduces pure competition over short‑term personal gain.

Taken together, these elements create a credible path for converting an installed technology investment into routine value — but only if safeguards, measurement design and role redesign follow through.

---

Clear and present risks: where incentives can go wrong​

While the mechanics are neat on paper, the Shoosmiths experiment also surfaces structural risks that any firm should consider before mimicking the approach.

• Metric gaming and hollow adoption
  Counting prompts is easy to track but easy to game. If rewards are tied to counts rather than validated impact, staff may issue low‑value or repeated prompts that improve telemetry but not client outcomes. Programs that measure raw usage without human verification risk creating vanity metrics.
• Data protection and confidentiality hazards
  Legal work routinely touches privileged, confidential or regulated information. Even when Copilot is used for non‑legal assistance, poor redaction, mistaken paste‑ins or misconfigured connectors can leak sensitive data into model logs. Incentives that encourage more use can amplify this risk unless strict technical and behavioural controls are in place.
• Apprenticeship erosion and distributional harm
  Automating routine drafting and review tasks can remove on‑the‑job learning opportunities for junior lawyers. If incentives accelerate automation without concurrently redesigning training and rotations, the firm may hollow its future talent pipeline. Empirical labour analyses suggest entry‑level roles are among the most exposed to early generative‑AI displacement.
• Surveillance perceptions and privacy backlash
  Telemetry tied to pay raises concerns about workplace surveillance. Even anonymised dashboards can feel intrusive if staff believe usage data will feed performance management decisions. Miscommunication or lack of consent on monitoring can damage trust.
• Regulatory and liability exposure
  The legal profession’s obligations — client confidentiality, competence and supervision — impose a lower tolerance for automation errors. Courts and regulators have shown a willingness to scrutinise filings that rely on unverified AI outputs. Incentive schemes must therefore be conservative about what activities are eligible for credit.

These risks are not hypothetical; they are practical, foreseeable, and in many cases already observed in early pilots and academic research. Shoosmiths acknowledged the tension and emphasised outcome quality over raw counts, but operationalising that rhetoric is the difficult part.

---

Designing better AI incentive programs: a practical playbook​

For firms and HR leaders considering a similar approach, the evidence suggests a set of design principles to capture upside while reducing harm.

• Tie rewards to validated outcomes, not just telemetry
• Reward verified time saved, template reuse across teams, or reductions in partner review time following human verification.
• Use juried or peer‑review panels for high‑value claims.
• Make rewards conditional on governance completion
• Completion of data‑handling and hallucination‑detection training should be a precondition for telemetry to count.
• Limit credit to use within sanctioned tenant instances and deny credit for public model usage.
• Reward reuse, sharing and diffusion, not raw volume
• Incentivise the creation of reusable prompts, verified templates and training sessions led by “power users.” This creates spillover benefits and reduces gaming incentives.
• Protect apprenticeship pathways
• Reinvest a portion of automation savings into structured rotations, mentorships and supervised drafting assignments to preserve learning opportunities.
• Separate telemetry from performance management
• Use anonymised dashboards for leadership and population‑level metrics rather than individualised surveillance used in promotion decisions. Obtain explicit consent where individual monitoring occurs.
• Run short pilots, measure, then scale
• Start with 6–12 week sandboxes, instrument both usage and impact metrics, and require human validation before scaling incentives firm‑wide.
• Auditability and logging
• Maintain exportable prompt logs, model version metadata and verification trails to satisfy regulatory or client scrutiny if contested. Ensure proper retention and deletion policies to meet privacy obligations.

These steps transform cash from a blunt force instrument into a targeted accelerant that buys auditable, high‑quality change rather than short‑term vanity metrics.

---

Operational guardrails for IT and compliance teams​

The technical and governance details matter. Practical controls to reduce risk include:

• Tenant lockdown and connector controls: only sanctioned connectors should be permitted; block or monitor file‑sharing with public model endpoints.
• Role‑based access and redaction tooling: ensure junior staff cannot inadvertently submit privileged data; provide easy redaction workflows.
• Hallucination detection and verification workflows: require human review for outputs used in client deliverables and create “AI verifier” roles where needed.
• Privacy notices and consent for telemetry: staff should be informed what is logged and how it is used; anonymise where possible.
• Incident response and audit trails: document and rehearse processes for data leakage or hallucination incidents tied to Copilot outputs.

The intersection of IT controls, legal duties and HR policy is where these programmes live or die. Tight technical governance paired with transparent people practices is required to avoid unintended consequences.

---

What Shoosmiths’ move signals for the legal market​

Shoosmiths’ public experiment matters beyond the size of the pot because it converts a boardroom AI commitment into a visible, profession‑level experiment. The programme does three things for the market:

• It signals to clients and competitors that the firm intends to embed AI into routine workflows rather than treat it as a novelty.
• It provides a large, auditable data set for other firms to study: does prompt usage translate into validated time savings, error reduction, or client satisfaction improvements?
• It forces a profession‑wide conversation about apprenticeship, regulation, and compensation design in an AI‑augmented practice.

If the programme yields durable client benefits while preserving training pathways, it could become a template for broader adoption. If it produces hollow metrics or harms junior development, it will likely be studied as an avoidable misstep. Either way, it accelerates the market’s learning curve.

---

Measuring whether the experiment “worked”​

Judging success requires moving beyond raw counts to three pillars:

• Quality and compliance — Did AI‑assisted outputs reduce errors or increase rework due to hallucinations? Track compliance incidents and client complaints attributable to AI use.
• Talent and learning — Did junior staff retain or improve drafting competence? Monitor promotions, attrition, and training completion among early‑career cohorts.
• Client value — Did validated time savings lead to faster delivery, higher client satisfaction, or measurable revenue gains? Use peer‑validated time‑saved metrics and client surveys to confirm.

Concrete KPIs should include verified hours saved per task, number of reusable templates adopted firmwide, partner review time pre‑ and post‑pilot, and distributional impact on entry‑level hiring and promotions.

---

Final verdict: conditional thumbs‑up, not a green light​

Shoosmiths’ one‑million prompts/£1 million experiment is consequential because it moves the debate from theory to measurable practice. As an accelerator for habit formation, it is smart: relatively small nudges can unlock large diffusion effects when paired with training and knowledge sharing. The collective design also mitigates purely selfish gaming.
However, the headline must be read alongside clear caveats: telemetry is an imperfect proxy for client value; the detailed mechanics of payout and timing remain less visible in the public record and deserve cautious interpretation; and the programme’s longer‑term value depends on governance, apprenticeship protection and rigorous validation of outcomes.
For firms contemplating similar incentives, the message is simple: use cash as an accelerant, not a substitute for governance, training and role redesign. Reward validated outcomes and reuse, protect apprenticeship pathways, and instrument both usage and impact before writing the cheque. When those pieces are present, incentives can convert platform spend into auditable change. When they’re missing, incentives buy visibility but risk structural harm.

---

Shoosmiths’ experiment will be watched closely by rivals, clients and regulators. If it demonstrates measurable client value without eroding professional standards, the legal market may see more creative, behaviour‑led adoption incentives. If it produces hollow metrics or unwanted distributional effects, the case will be cited as a cautionary tale. Either outcome sharpens the profession’s understanding of how to integrate Microsoft Copilot, enterprise AI, and new human‑machine workflows into the craft of law.

---

Source: Non-Billable Shoosmiths hits AI target, unlocking £1m bonus pool