THE EMPATHY PREMIUM IN 2026 CORPORATE RESTRUCTURING

Why Human Leadership Still Matters.

Corporate restructuring in 2026 has somehow managed to make bad news look immaculate. The earnings call is polished, the synergy spreadsheet is wearing its best tie, and somebody has suggested that artificial intelligence deliver the announcement because it “removes emotion from the process”. That suggestion usually comes from someone who won’t be in the room when the process removes a salary, medical aid and what remains of the workforce’s faith in management.

AI can identify duplicated roles, model costs and find patterns in alarming quantities of data. It can draft a consultation timetable before the CFO has finished saying “optimisation”. What it can’t do is carry moral responsibility, read the silence after an announcement or repair trust when employees suspect the decision was made by a machine nobody can explain.

That gap is the empathy premium: the commercial, legal and cultural value created when leaders combine sound data with judgement, dignity and accountable human conversation. During layoffs, mergers and crises, empathy isn’t scented stationery from human resources. It’s risk management with a pulse.

Corporate Restructuring in 2026: The Spreadsheet Is Not the Company

The temptation is obvious. Business needs a faster answer. AI produces one, complete with percentages and the serene confidence of a consultant who has already invoiced. Yet the International Labour Organization’s 2025 update found that one in four workers globally is in an occupation with some generative-AI exposure, while stressing that transformation, rather than wholesale redundancy, is the more likely outcome because most jobs still require human input. The sensible question is therefore not, “How many humans can we delete?” It is, “How should work be redesigned, and who bears the consequences?”

In South Africa, that question lands in a labour market shaped by inequality, unemployment, scarce skills and long institutional memories. A restructuring is rarely experienced as an elegant organisational redesign. It’s experienced, rather heart-breakingly, at kitchen tables. The corporate euphemism may be “rightsizing”, but school fees remain stubbornly full-sized.

South African research makes the point less theatrically. A study of retained employees in a restructuring mining organisation found that improved supervisor support enhanced psychological safety and work engagement, with psychological safety mediating that relationship. Separate South African research found a strong relationship between restructuring, institutional trust and employee engagement. Apparently, workers perform better when leadership behaves like leadership. A shocking development.

AI Layoff Decisions Create a Human Trust Deficit

A model can rank roles against selected variables, but the ranking is only as wise as the assumptions fed into it. The payroll clerk labelled “administrative duplication” may be the only person who knows why three legacy systems disagree every month. Removing one multilingual client-facing employee may also damage relationships across several communities. Unless people design, test and challenge the system properly, a low metric can quietly become a proxy for structural disadvantage.

The OECD’s 2026 survey of more than 6,000 firms records familiar concerns about algorithmic management: unclear accountability, difficulty following a tool’s logic and inadequate protection of workers’ health. Those are awkward qualities in any management system. In a retrenchment exercise, they’re the sort of qualities that arrive later wearing a counsel’s robe.

Fairness also has several dimensions. Research on layoff survivors highlights distributive, procedural, interpersonal and informational justice. Employees assess not only what happened, but how decisions were reached, how the news was communicated and whether they were treated with respect. Empathetic, transparent practices can preserve trust. An automated message beginning “Dear valued resource” is unlikely to do the same.

South African Retrenchment Law Requires Consultation, Not Algorithmic Theatre

Section 189 of the Labour Relations Act 66 of 1995 requires a meaningful consultation process when dismissals for operational requirements are contemplated. The process addresses ways to avoid dismissals, minimise their number, change timing, mitigate adverse effects, select employees fairly and determine severance pay. Consultation isn’t a ceremonial webinar held after the answer has been locked in a dashboard. It’s a joint problem-solving exercise.

The Department of Employment and Labour’s guidance describes retrenchment as a last resort and identifies the governing statutory framework. That matters because a model may optimise for speed while the law insists on process. “The computer has already decided” isn’t consultation. It’s an admission with graphic design.

The discrimination risk is equally serious. Section 6 of the Employment Equity Act 55 of 1998 prohibits direct and indirect unfair discrimination in employment policies and practices on listed and arbitrary grounds. A selection model trained on historical promotion, attendance, pay or performance data may reproduce patterns linked to sex, pregnancy, family responsibility, age, disability, race or other protected grounds. Bias doesn’t become objective because it has an interface.

Human employment lawyers are therefore needed before, during and after model deployment – to test the business rationale, interrogate proxies, examine adverse impact, protect privilege, structure consultation and insist on records that explain who decided what. The lawyer’s job is not to sprinkle legal dust over an automated outcome. It’s to stop efficiency becoming evidence.

Employment Lawyers Add Emotional Intelligence to Executive Severance

Executive exits expose the limits of pure computation. A severance package may involve notice, incentives, restraints, confidentiality, intellectual property, board duties, regulatory disclosure, tax, reputation and the choreography of an announcement. The numbers matter. So do status, identity, fear and the powerful executive desire to leave “by mutual agreement” shortly after discovering the agreement wasn’t entirely mutual.

A skilled human lawyer hears what isn’t being said. They can recognise when money is a proxy for dignity, when an apology may unlock settlement, when a public statement will inflame the dispute and when an apparently small drafting concession protects both sides. AI can compare clauses. It can’t own the room, absorb anger or make a credible judgement that tomorrow’s headline is worth more than today’s saving.

This isn’t an argument for mystical lawyering or expensive handholding. It’s an argument for accountable discretion. The lawyer must integrate evidence, legal principle, organisational context and human behaviour, then stand behind the advice. A machine doesn’t lose sleep over a bad recommendation. It simply awaits the next prompt, refreshed and legally untroubled.

Mergers Fail When Leaders Automate Culture

Mergers are sold in the language of synergies, platforms and scale. Employees hear – Which system survives? Which office closes? Whose title disappears? McKinsey’s research reports that cultural differences and changed operating models account, on average, for almost half of mergers that fail to meet expectations. Culture isn’t the soft part after the deal. It’s where the deal either becomes real or quietly starts eating itself.

AI can map reporting lines and identify duplicated licences. It can’t reconcile two professional identities, notice which rituals hold a team together or explain why changing an expense policy feels like cultural occupation. Leaders must name uncertainty, answer difficult questions repeatedly and make visible choices about whose practices continue. The phrase “we’re creating one culture” isn’t a strategy. It’s usually what appears on slide 14 just before everyone returns to defending their old culture.

How Legal Tech Suppliers Can Build the Empathy Premium

As a legal tech supplier, we’ve learnt that the strongest systems don’t replace professional judgement, they create the time and evidence needed to exercise it properly. We’re working to make human review, challenge and explanation visible wherever a decision carries serious consequences. That means resisting the seductive promise that a model should decide without interference. In our experience, that isn’t innovation. It’s a future exhibit.

Our practical challenge going forward is to build technology that strengthens the professionals using it rather than staging a hostile takeover of their judgement. The more consequential the decision, the clearer we must make the points at which a person can question, override and explain the output. We’re not pretending we’ve solved every part of that challenge. We are, however, treating accountable human oversight as part of the product rather than an inconvenient interruption to it.

Keep humans accountable for employment decisions. Use AI to organise evidence, model scenarios and surface anomalies, but require an authorised person to interrogate assumptions, consider individual context and record reasons before any adverse decision. Human-in-the-loop must mean authority to disagree, not permission to admire the output before approving it. Audit the data for South African discrimination risk. Test inputs and outcomes for direct and indirect disadvantage, including effects linked to disability, maternity, family responsibility, pay history, language, geography and access to work. “The dataset was large” isn’t a defence if it was largely wrong. Finally, design consultation before designing communication. Give affected employees and representatives enough information to engage meaningfully, test alternatives and reach the people who actually hold decision-making authority. A chatbot can collect questions. It cannot convert a predetermined outcome into good faith.

In addition, keep final conversations, negotiations, appeals and crisis decisions human. The point of legal technology is to give professionals more capacity for judgement, not more sophisticated ways to avoid it.

Global AI Regulation Is Moving Towards Human Oversight

South African employers must start with South African law, but global developments show the direction of travel. The European Union’s AI Act uses a risk-based framework and treats specified employment applications as high risk, with requirements including risk management, documentation, transparency and human oversight. The revised application timetable for employment-related high-risk rules extends into 2027, but the governance signal is already clear: consequential automation requires controls.

Colorado’s 2026 automated-decision law, effective in 2027, similarly covers consequential employment decisions and gives affected people rights to notice, correction and meaningful human review after adverse outcomes. Jurisdictions differ and imported compliance slogans should never substitute for local advice. Still, the international trend is not towards “the algorithm said so”. It’s towards explanation, accountability and review.

For South African legal tech suppliers, this is an opportunity. Build products that preserve reasons, disclose limitations, enable bias testing and make review visible. Sell trustworthy infrastructure rather than synthetic omniscience. Clients need better instruments, not a robot oracle with a subscription tier.

Legal Technology Should Create Time for Human Advocacy

This is where legal technology earns its place. Practice-management, billing, accounting, workflow and document systems can reduce friction, strengthen record-keeping and free professionals from administrative sludge. That creates time for the work machines cannot responsibly own: listening, advising, negotiating, challenging assumptions and carrying accountability.

The Empathy Premium Is a Corporate Asset, Not a Courtesy

The organisations most likely to survive volatile transitions are not those that reject AI. They’re the ones that know where its usefulness ends. Technology can sharpen the evidence, but human leaders and employment lawyers must still make the difficult decisions, explain them honestly and take responsibility for what follows.

The empathy premium appears in fewer disputes, stronger consultation, retained knowledge, better integration and the willingness of surviving employees to believe management again. Those outcomes may not fit neatly into the first restructuring model. Neither does the cost of losing them.

Let AI balance the ledger, compare the clauses and schedule the meetings. When livelihoods, dignity, and trust are on the line, put a competent human in the room. Preferably one who has read the law, checked the data and knows that “Dear valued resource” isn’t an empathy strategy.

---

If your firm is trying to automate the administrative burden without automating away judgement, dignity or accountability, talk to AJS. We’ll help you explore practical legal technology that supports your people, strengthens oversight and gives professionals more time for the work that still needs a human in the room. Let’s start with the problem you’re actually trying to solve – not the product somebody is trying to sell you.

– Written by Alicia Koch on behalf of AJS

(Sources used and to whom we owe thanks:  Labour Relations Act 66 of 1995;  Retrenchment and Your Legal Rights; Employment Equity Act 55 of 1998;  Heyns, M.M., McCallaghan, S. and Senne, O.W. (2021). Supervisor support and work engagement: The mediating role of psychological safety in a post-restructuring business organisation; Marais, A. and Hofmeyr, K. (2013). Corporate restructuring: Does damage to institutional trust affect employee engagement?  Generative AI and jobs: A 2025 update; Algorithmic management in the workplace; Lee, S., Hong, S., Shin, W.-Y. and Lee, B.G. (2023). The Experiences of Layoff Survivors: Navigating Organizational Justice in Times of Crisis; Managing and supporting employees through cultural change in mergers; Artificial Intelligence Act: regulatory framework and SB26-189 Automated Decision-Making Technology)