Skip to content

Vennlisk and AI: what the case study says and what it doesn't

What the Vennlisk SCS case says about AI and chatbots, what it leaves out, and how that shapes answers on strategy, risk and staff.

SCSPre-seen insights

Download CIMA's pre-seen from the CGMA Study Hub and read it alongside this summary.

AI runs through the Vennlisk pre-seen, but it is easy to read more into it than is there. This article separates what the pre-seen says from what it doesn't, and then gives our reading of what that means for your answers. Download CIMA's pre-seen from the CGMA Study Hub and read the pages cited here; page numbers are the pages of CIMA's PDF. For the case as a whole, see our main Vennlisk article.

What the pre-seen says

Vennlisk is evaluating AI, not running it. The pre-seen says Vennlisk is evaluating AI software that could support and manage agents' work (p12). It gives three kinds of use:

  • Help during a call. Summarising a caller's file so the agent can reach a good outcome, for example by suggesting accessories to someone buying a product from an online retailer (p12).
  • Watching for stress. Monitoring conversations for signs of stress and flagging concerns to the supervisor, who can listen in and step in if needed (p12).
  • Reading the metrics. Spotting patterns in agents' performance, such as an agent whose sales rise while customer satisfaction falls, which could lose business over time (p12).

Some automation already exists. Among the mitigations for its principal risks, the pre-seen says some systems use interactive voice response, letting callers complete simple tasks without an agent so agents can focus on more complex work (p17). The industry section also notes that software can track customer satisfaction during a call from tone of voice or keywords (p8); it describes that as something call centres can do, not something Vennlisk does.

A press article weighs up chatbots. A business news piece describes chatbots as AI software that holds text or voice conversations (p26). It sets out the case for them (they can replace human assistance and run around the clock) and against (they often miss mood and emotion, struggle with complex conversations, and can give credible but wrong advice if their training data isn't kept current and checked).

Innovation is in the values. Innovation appears in Vennlisk's vision and values (p13), and "innovation and technology" is one of the five areas its sustainability section names (p23).

The people. The chief executive, Chen Yue, is an electronic engineer specialising in telecommunications who joined Vennlisk to run network administration (p13). The Operations Director, Dr Maria Chandler, is responsible for call centre operations and software development (p15).

What the pre-seen doesn't say

Knowing the gaps stops you inventing facts, and tells you what a trigger might add. The pre-seen gives:

  • No AI in live use beyond interactive voice response.
  • No budget, supplier, timetable or pilot for the evaluation.
  • No plan to replace agents. Every example has AI helping agents and supervisors.
  • No client demand. Nothing says clients are asking for AI or automating their own calls.
  • No competitor activity. Nothing says Ringcoll or overseas centres are using AI.
  • No AI rules. Nothing describes how Westland regulates AI or automated calls.
  • No director with a stated AI background. No biography mentions AI or data science (pp13–15).

If a task gives you one of these, it is new information, and your answer should be built around it.

Our reading: what it means for the strategy

Pricing decides who benefits. Most clients pay Vennlisk an hourly rate for dedicated agents; others pay per minute of agent time (pp7, 11). Under either model, a tool that makes calls shorter cuts the hours or minutes Vennlisk bills. In our reading, efficiency gains from AI would flow largely to clients unless Vennlisk changes how it charges, for example towards fees per resolved query. That connection between technology and the business model is where strong SCS answers tend to start.

AI fits the quality strategy, up to a point. The board's stated strategy is a good quality service that clients value despite cheaper overseas rivals (p27). The pre-seen's examples, better-informed agents and quicker help for stressed callers, support that strategy. A move to chatbots answering calls end to end would cut against it, and the chatbot article's limits (emotion, complexity, credible errors) are exactly where Vennlisk claims strength (pp12, 26).

Paying for it. The pre-seen gives no cost for AI, but it does show the constraints: borrowings of W$5,460.0m, interest cover of about 3.3 times and a dividend of W$509.6m against profit of W$770.8m (pp18–19). In our reading, a large AI programme would force a choice between the dividend, new funding and a slower roll-out. See Vennlisk against Ringcoll for how its rival compares.

Our reading: risks and people

Privacy. Access to client files is already a principal risk (p17). AI that reads callers' files, or that is trained on recorded calls, adds to it. Controls over what data an AI tool sees, where it is processed and what clients' contracts allow belong in any recommendation.

Credible wrong answers. The chatbot article's warning (p26) applies to agent-assist tools too: a confident but wrong suggestion on screen could mislead an agent. Human checks, approved knowledge sources and monitoring of outcomes are the obvious controls.

Monitoring staff. Flagging stress and tracking performance patterns could help agents, but it is also surveillance. Agents are already tracked closely (p12), and a new agent's blog describes pressure from a daily call count (p25). In our reading, how staff are consulted and told, and whether monitoring is used to support or to discipline, is an ethical question as much as an operational one.

Skills on the board. With no director whose background is in AI, a task could ask how the board would oversee an AI programme: training, an adviser, a new non-executive or a committee brief. The Risk and Sustainability Committee (p16) is the existing home for oversight of risk.

Jobs. The pre-seen's examples don't remove agents, but interactive voice response already moves simple tasks away from them (p17). If a trigger proposes cutting agent numbers, the stakeholders include agents, supervisors, the communities around five city centres, clients and shareholders. Consultation, timing and honesty about the reasons are part of a Level 3 answer.

How to use this in the exam

  • Start from the trigger. If the unseen adds a supplier, a cost or a client request, build your answer on it rather than on general AI benefits.
  • Link AI to pricing and quality. Who pays, for what, and who keeps the saving? Does the proposal support or undermine the quality strategy?
  • Weigh it, then recommend. Benefits, risks, cost and staff impact, then a clear recommendation with its main risk and a next step, such as a pilot with one willing client.
  • Use the syllabus. Generative AI and digital strategy sit in E3; AI-enabled fraud and data risk in P3; funding and evaluating the investment in F3.

Every figure mentioned here, with its page, is in the figures pack on our Vennlisk case study page. For the other themes in the case, see our main Vennlisk article.

More insights