You are viewing a free preview of this lesson.
Subscribe to unlock all 13 lessons in this course and every other course on LearningBro.
Spec mapping: AQA 7138 Unit 3.3.2 — Business and the External Environment (refer to the official AQA specification document for exact wording). This lesson develops AI as a disruptive technology at A-Level depth — AI's status as a general-purpose technology (Brynjolfsson framing) that diffuses across the economy in a way comparable to electricity or the internal-combustion engine, Christensen's disruption framework as applied to AI-driven competitive dynamics, the labour-market impacts (augmentation vs displacement), the EU AI Act's risk-tiered regulatory framework, the data-ethics dimension (algorithmic bias, accountability, transparency, AI safety), and the analytically loaded question of which AI-integration strategy a professional-services firm should adopt as AI capabilities mature. The 15-mark Evaluate on this lesson is the second discriminator tariff for the batch — does the candidate construct two genuinely contestable AI-strategy options, deploy multiple Annex 8 sophisticated concepts with conceptual rigour, and reach a defended on-balance judgement that explicitly weighs the augmentation-vs-displacement and strategic-drift dimensions?
Connects to:
The conceptual frame for understanding AI's economic significance is the general-purpose technology (GPT) framework developed in the economic-history literature. A general-purpose technology is one that diffuses across the entire economy, enables complementary innovations in many sectors, and reshapes productive arrangements over a multi-decade horizon. Historic examples include the steam engine (18th–19th century), electricity (late 19th–early 20th century), the internal-combustion engine (20th century), and the internet (late 20th–early 21st century).
Erik Brynjolfsson and collaborators have argued that contemporary AI — particularly large-language-model and machine-learning systems — exhibits the GPT pattern. The defining features are: (i) pervasiveness — AI applications are emerging across nearly every sector; (ii) innovation-enabling — AI is a platform for further innovation, not just a single-purpose tool; (iii) complementary investment required for productivity gains — businesses must invest in data infrastructure, workforce reskilling and process redesign for AI investment to deliver returns.
A specific Brynjolfsson framing is the productivity J-curve — the observation that GPT-level technologies typically deliver disappointing measured productivity in their early diffusion phase before delivering substantial productivity gains as complementary investments mature and process redesign catches up with technological capability. The J-curve framing implies that businesses making early AI investments may experience near-term productivity disappointment before substantial gains arrive; businesses waiting for proven AI productivity may find themselves on the wrong side of the curve as competitors that invested early move ahead.
The conceptual significance for strategic decision-making is that GPT investment cannot be evaluated through conventional short-payback investment-appraisal because the productivity gains arrive on a multi-year horizon and depend on complementary investments that extend the apparent payback period. Conventional NPV (Annex 7 formula 51) and payback (Annex 7 formula 49) analysis tends to under-state GPT investment value.
Clayton Christensen's disruption framework (1997 onwards) provides one of the most influential lenses for understanding how technological change reshapes industry structure. The framework distinguishes sustaining innovation (incumbents-led improvements that serve mainstream customers better) from disruptive innovation (initially-inferior technology that serves under-served or non-consumer segments before moving up-market to challenge incumbents).
Applying the Christensen framework to AI requires distinguishing between AI applications that are sustaining (incumbents-led automation of existing processes — for example, AI-augmented document review by major law firms or AI-driven trading by major investment banks) and AI applications that are disruptive (initially-inferior services that target under-served segments and move up-market — for example, AI-only legal-document services targeting consumer and small-business markets, AI-driven robo-advice in financial services).
The exam-relevant analytical move is that disruptive AI threatens incumbent positions more than sustaining AI. Incumbents that respond to AI primarily through sustaining-innovation framing (using AI to make existing processes more efficient) may miss the disruptive-innovation threat from below. The classic Christensen failure pattern is incumbents over-serving the high-end of their existing market while disruptors capture the lower end and move up.
The specific incumbent's dilemma in AI strategy is that integrating AI into existing service offerings may cannibalise current revenue (each AI-augmented junior-lawyer hour bills less than the un-augmented hour it replaces) while not investing in AI surrenders ground to disruptive entrants. The dilemma is most acute for high-margin professional-services businesses whose existing economics depend on selling labour-intensive expertise at high hourly rates — AI both threatens the existing economics and creates new service-design opportunities that the incumbent may be structurally slow to capture.
The labour-market effects of AI are the most-debated dimension of AI's economic significance. The conceptual framing distinguishes augmentation (AI tools that make existing workers more productive, with workers retained in modified roles) from displacement (AI systems that replace human workers, with reduced headcount in affected roles).
| Pattern | Description | Examples |
|---|---|---|
| Pure augmentation | AI tool extends worker capability; worker retained in modified role | AI-augmented code completion; AI-augmented medical-imaging interpretation |
| Selective displacement | AI replaces some workers in a role while others are retained for higher-complexity work | Customer-service chatbots displacing tier-1 support while tier-2 support is retained |
| Role disappearance | An entire role category contracts substantially as AI substitutes for it | Some categories of routine office support, document processing, content moderation |
| Role creation | New roles emerge around AI development, integration, governance and oversight | Prompt engineers, AI-ethics officers, AI-systems product managers |
A specific A-Level analytical move is to recognise that the labour-market impact of AI is deeply uncertain in the technical sense — the magnitude of displacement, the timing of substitution, and the speed of role-creation are all subject to forecasting error far larger than conventional economic-cycle uncertainty. The risk vs uncertainty concept (Annex 8 analytical concept #d10) is the analytical anchor — the labour-market consequences of AI cannot be reliably modelled from historical analogies (because the technology is genuinely novel) or from current diffusion-rate extrapolation (because complementary-investment and adoption dynamics are non-linear).
The strategic implication is that workforce-planning under AI requires option-value thinking rather than expected-value optimisation. Businesses that build flexibility into workforce design (reskilling capability, role-redesign capacity, ability to scale specific functions up or down) preserve strategic optionality; businesses that commit to single workforce configurations bet on specific AI-trajectory forecasts that may not hold.
The EU AI Act (adopted 2024, phased implementation from 2025–2027) is the world's first comprehensive AI regulation. Its architecture is risk-tiered — different regulatory requirements apply depending on the assessed risk level of the AI system.
| Risk tier | Description | Regulatory treatment |
|---|---|---|
| Unacceptable risk | AI uses that fundamentally conflict with EU values | Prohibited (e.g. social-scoring by governments, certain biometric surveillance) |
| High risk | AI uses in critical contexts (employment, education, law enforcement, critical infrastructure) | Conformity assessment; transparency, accuracy, robustness and human-oversight requirements |
| Limited risk | AI systems interacting with humans (chatbots, generated content) | Transparency obligation (users must know they are interacting with AI) |
| Minimal risk | General-purpose AI applications | No specific AI-Act obligations (general consumer-protection still applies) |
The Act includes general-purpose AI model (GPAI) provisions that impose additional transparency, safety and governance requirements on foundation-model providers, with stricter requirements for models above specified compute and capability thresholds.
The UK has taken a pro-innovation regulatory approach that differs from the EU framework — the UK government has signalled preference for sector-specific regulator-led oversight (FCA, ICO, MHRA, Ofcom each applying AI principles within their remit) rather than horizontal AI-specific legislation. The UK position remains under development, and the practical effect for UK businesses depends partly on whether they operate in EU markets (and therefore face EU AI Act compliance for their EU-facing operations).
The strategic implication for UK businesses is dual-compliance complexity — UK businesses with EU operations face the EU AI Act's risk-tier requirements alongside the UK's sector-regulator approach. This is structurally similar to the post-Brexit regulatory-divergence pattern in other regulated sectors.
The data-ethics dimension of AI strategy captures the substantive ethical concerns about AI deployment that interact with regulatory requirements and stakeholder expectations.
AI systems trained on historical data can reproduce and amplify historical biases in their predictions and decisions. Hiring-screening AI trained on past hiring data may reproduce gender or ethnicity biases; lending-decision AI trained on past lending data may reproduce racial-or-geographic biases. The conceptual significance is that AI does not eliminate human bias; it operationalises the bias of the training-data generators and embeds it into systematic decision-making at scale.
The technical responses include bias auditing (testing AI outputs for disparate impact across protected characteristics), training-data curation (constructing training data to balance under-represented groups), and outcome monitoring (tracking AI-decision distributions for emergent bias patterns). The technical responses do not eliminate bias risk but reduce it.
The accountability concern is that AI decisions affecting individuals (hiring, lending, insurance, healthcare access) should be explainable and contestable by those affected. Many AI techniques (deep neural networks particularly) produce decisions whose internal reasoning is opaque even to their designers. The accountability concern has driven the development of explainable AI techniques that produce decision-rationales alongside decisions, with mixed success.
The transparency concern is that users interacting with AI systems should know they are doing so. The EU AI Act's limited-risk-tier transparency obligation operationalises this concern through disclosure requirements for chatbots and AI-generated content.
The AI safety concern (most prominent in the foundation-model context) is the broader question of whether sufficiently capable AI systems can be reliably aligned with intended outcomes and prevented from producing harmful behaviour. AI-safety concerns range from near-term issues (misuse of generative AI for fraud, disinformation, malware) to longer-term existential concerns about advanced AI systems. The UK government has been particularly active in AI-safety policy, hosting AI-safety summits and establishing the AI Safety Institute.
flowchart TD
GPT["AI as general-purpose<br/>technology (Brynjolfsson)"] --> Diffusion["Cross-sector diffusion"]
Disruption["Christensen disruption:<br/>sustaining vs disruptive"] --> Choice["Strategic AI choice"]
Diffusion --> Choice
Choice --> Labour["Labour-market impact:<br/>augmentation vs displacement"]
Choice --> Regulation["Regulatory landscape:<br/>EU AI Act / UK sector"]
Choice --> Ethics["Data ethics:<br/>bias / accountability /<br/>transparency / safety"]
Labour --> Workforce["Workforce strategy:<br/>reskilling / redesign /<br/>optionality"]
Regulation --> Compliance["Compliance design:<br/>risk-tier assessment /<br/>dual-jurisdiction"]
Ethics --> Trust["Stakeholder trust:<br/>customers / employees /<br/>regulators"]
Workforce --> Outcome["Strategic outcome:<br/>competitive position /<br/>strategic drift risk"]
Compliance --> Outcome
Trust --> Outcome
style GPT fill:#1d4ed8,color:#fff
style Choice fill:#a16207,color:#fff
style Outcome fill:#15803d,color:#fff
style Ethics fill:#b91c1c,color:#fff
The diagram captures the integrated logic — AI's GPT character and Christensen-disruption dynamics shape the strategic AI choice, which generates downstream consequences across labour-market, regulatory and ethical dimensions, each of which feeds into competitive-position outcomes and strategic-drift risk.
Holborn Stratton LLP is a hypothetical UK mid-market professional-services firm (legal, tax-advisory and corporate-finance services), established 1987, with 1,240 fee-earners and 480 support staff across six UK offices and two EU offices (Dublin, Amsterdam). 2025 revenue was £498 million; operating profit margin 19.7 %; average partner profit share approximately £580,000. The firm's competitive position rests on senior-partner expertise, blue-chip corporate-client relationships and a junior-talent pipeline drawing 110 trainees annually from UK and EU law schools. The firm faces a strategic-AI decision as foundation-model and legal-AI capabilities mature rapidly. Two options are under consideration: Option A — defensive automation (deploy AI tools to automate existing junior-lawyer and support-staff tasks — document review, due-diligence checklists, contract-clause comparison, time-sheet processing; reduce trainee intake by 40 %; capture cost-savings within current service-pricing model); Option B — service-model reinvention (redesign the firm's service offerings around AI-augmented expertise — fixed-fee AI-enabled legal-and-tax compliance products for mid-market clients, AI-driven scenario modelling for corporate-finance work, productised legal-research subscriptions; retain trainee intake to support new service-design capability; invest £18 million over 36 months in AI platform, change management and product design). EU AI Act compliance for the firm's Dublin and Amsterdam operations is in scope under both options.
Figures and company are fabricated for illustrative purposes; not affiliated with any actual business.
Evaluate which of the two AI-strategy options Holborn Stratton should adopt. (15 marks)
| AO | What the question rewards | Mark weighting on this 15-mark item |
|---|---|---|
| AO1 | Knowledge of AI as general-purpose technology, Christensen disruption framework, augmentation-vs-displacement labour-market dimension, EU AI Act risk-tier framework, data-ethics dimension, strategic-drift risk | ~3 marks |
| AO2 | Application to Holborn's specifics — £498m revenue, 19.7 % operating margin, 1,240 fee-earners, partner-profit-share structure, trainee-intake model, six UK + two EU offices, professional-services business model | ~3 marks |
| AO3 | Analytical chain-of-reasoning — what does the partner-profit-share structure imply for the incumbent-dilemma framing? How does the trainee-pipeline decision interact with future-capability-build? Why does EU AI Act compliance matter for both options? | ~4 marks |
| AO4 | Evaluation judgement — does the strength of the defensive-automation case (immediate cost-saving, manageable change, preserved current economics) outweigh the strength of the service-model-reinvention case (disruptive-threat protection, productised-service opportunity, future-capability-build), given Holborn's specific position? Deploys ≥2 Annex 8 sophisticated concepts. | ~5 marks |
15-mark Evaluate items reward a structured propose-and-evaluate build with a defended on-balance judgement. Annex 8 sophisticated-concept deployment is the discriminator between Stronger-band and Top-band.
AI is a major external change affecting professional-services firms. Holborn faces a choice between Option A (defensive automation) and Option B (service-model reinvention).
Option A is attractive because it captures immediate cost-savings within the current business model. Automating junior-lawyer tasks and reducing trainee intake by 40 % delivers near-term margin improvement. The change is more manageable than a full service-model redesign, and current partner-economics are preserved.
Option B is attractive because it positions Holborn for the disruptive-threat that AI poses to traditional professional-services. Christensen's disruption framework suggests that incumbents who only optimise existing processes may miss the disruptive innovation from below. The £18m investment is significant but builds future capability in AI-enabled service design, productised legal services and AI-augmented expertise.
The risk of Option A is strategic-drift — if Holborn only automates existing tasks while competitors reinvent service models, Holborn's competitive position erodes over time. The risk of Option B is the substantial upfront investment and the uncertainty about whether new service models will gain market traction.
On balance, Holborn should adopt Option B because the disruptive-threat from AI is real and the partner-profit-share economics depend on Holborn maintaining its position as a leading professional-services firm.
Examiner-style commentary: This response reaches Mid-band. AO1 references Christensen and strategic-drift; AO2 applies the key figures. The AO3 chain identifies relevant connections but does not push them deeply. To reach Top-band, the response needs to deploy multiple Annex 8 sophisticated concepts by name (risk vs uncertainty, stakeholder vs shareholder approaches, Carroll's CSR pyramid, strategic drift) and to develop the incumbent's dilemma framing in detail.
Subscribe to continue reading
Get full access to this lesson and all 13 lessons in this course.