How AI Changes the Issue-to-Resolution Value Stream

How customer service shifts from an agent handling every ticket to AI handling the routine ones with a person on call, and where that quietly goes wrong.

Sebastian Bitter

4.8.2026

Issue-to-Resolution carries a customer problem from the first message to a solved case. It is the stream where AI in service shows up first, and where getting it wrong is most visible to customers. As AI takes on more, the routine cases get handled automatically and the person moves to the hard, sensitive and high-value ones. The escalation path to a human becomes a designed step in the flow. For most companies the realistic target is AI resolving the standard cases with a reliable human escalation. Full automation without that escape hatch is where trust breaks. The clearest evidence is one company across two years. In its first month Klarna's assistant did the work of about 700 agents. A year later, after pushing to full automation, Klarna said quality had dropped and brought people back. A high resolution rate is not the same as a happy customer. Some vendor numbers count a case as solved when the customer simply goes quiet, so we read them with care.

The stream today, and where AI takes it

Issue-to-Resolution is the value stream that carries a customer problem from the first contact to a solved case. It starts when someone reports an issue and ends when the issue is resolved and the customer is told. Every company with customers runs this stream, and it is where a lot of today's AI attention in service lands, because the volume is high and the work repeats.

In its classical form the work runs as six steps. The customer gets in touch and an agent logs the ticket by hand. The agent sorts it by topic and urgency. The agent looks up the account history and the knowledge base. The agent writes and sends the answer. Hard cases go to a second-level team. Finally the agent closes the case and asks for a rating.

The pain is familiar to anyone who has waited on hold. A person does every step, the customer waits, and the same routine questions get answered again and again. Most of the delay sits in the queue and the hand-offs, not in the answer itself. Adding a faster search tool for agents does not fix that, because the bottleneck is how much a human team can handle at once.

This is what AI changes. Intake, sorting and the standard answers move to the AI. The person shifts to the cases that actually need judgment: the sensitive, the complex and the high-value ones. And the step where a case is handed to a human, the escalation, stops being an afterthought and becomes a designed part of the flow. The rest of this post walks that path one level at a time and shows each change with a real example.

Figure 1: Issue-to-Resolution across the four levels .The same stream at each level. Intake, triage and standard resolution move to the AI, and the human moves to the exceptions, with escalation as a designed gate.

Level by level, shown by real cases

Each level below comes with a real example, its headline number, and the point where that number stops holding. The classical level is the starting point above, with no AI in the flow, so we begin at assistive.

Assistive: AI drafts the reply, the agent sends it

At the assistive level the six steps stay as they are, and AI helps inside each one. It pre-fills the ticket, suggests a category, surfaces the account history and the right help article, and drafts the reply. The agent reviews every draft and sends it. The customer still talks to a person.

The example is Magic Ink, the support tool built by Kraken Technologies and used at the energy retailer Octopus Energy. It drafts replies from the customer's account history, and the company reported that about a third of those drafts were good enough to send with no changes. The agent stays in charge and sends every reply.

The help is real, but it has two weak spots. A draft that looks right is easy to send under time pressure, so the human check can weaken without anyone noticing. And the flattering numbers deserve a second look: the internal satisfaction score of 80 came from the easy cases, while the official public measure sat closer to 70. The tool helps agents on routine, high-volume mail. It struggles on the complex, multi-step cases, and those are the ones customers care about most.

Figure 2: The Kraken Magic Ink case. The customer message, the AI draft, the agent's decision and the sent reply. About a third of drafts went unchanged, on routine mail, with the agent sending every one.

Agentic: AI resolves the routine cases, a human is one step away

At the agentic level the stream is rebuilt around the AI, and the steps merge. Intake and sorting happen automatically. The AI then resolves the standard cases end to end, in the customer's own language: refunds, returns, invoice questions and the like. The important addition is a designed escalation gate. Sensitive or complex cases are routed to a human, and the customer can ask for a person at any point. The human becomes the exception handler and sets the policy.

The example is Klarna, the buy-now-pay-later company, and it is worth following across two years. In its first month Klarna reported that its OpenAI-based assistant handled 2.3 million chats, the equivalent of about 700 full-time agents, and cut the average resolution time from 11 minutes to under 2. The assistant did the routine work, and a customer who wanted a human could still reach one. These numbers come from Klarna itself, not an independent audit. This is the agentic level, and it worked.

A high resolution rate is not the same as a satisfied customer. One widely quoted support assistant, Intercom Fin, is advertised at a 76% resolution rate, while independent estimates put the real figure closer to 45 to 53%, partly because a case can be counted as resolved when the customer simply stops replying for a day. Read the rate together with whether the customer actually came back happy.

Figure 3: The Klarna arc, from agentic success to over-automation. One company across two years. In 2024 the AI resolves the standard case with a human one step away. In 2025 that escalation path is removed, quality drops, and Klarna brings people back.

Autonomous: full automation with no way out, and why it backfired

At the autonomous level the human escalation path is removed. The AI handles every case, and people only set the policy and watch the aggregate numbers. On paper this is the cheapest setup. In customer service it is also the riskiest, and the same Klarna case shows why.

A year after the success, Klarna pushed hard toward full automation and took the human escalation away. The company then said publicly that service quality had dropped and started bringing people back, on a flexible, on-demand basis. The missing escalation path was the root cause. When every case is forced through the AI and a frustrated customer cannot reach a person, the cost shows up in trust rather than in the support budget. The 2024 numbers were Klarna's own, and the 2025 walk-back is reported by independent press and a CEO admission against interest.

This is why the autonomous level in this stream is a warning. The same arc that shows the agentic win also shows the cost of pushing past it, so we keep both chapters in the one view above (Figure 3).

What is realistic

Put the levels together and the advice is clear. For most companies the realistic target in Issue-to-Resolution is AI resolving the standard cases with a reliable human escalation always available. The AI takes the routine volume, and anything sensitive, complex or high-value reaches a person quickly. That is the agentic Klarna pattern before the over-automation, and the escalation gate is the part you cannot skip.

Getting there takes a few things, and they are concrete. You need a clear playbook for the cases the AI is allowed to handle. You need a connection into your systems so it can actually do things like issue a refund, with the narrowest access that still works. You need clear rules for when a case goes to a human. And you need a structured knowledge base the AI can draw on. The cleaner those are, the more the AI can safely take.

There is a second cost that is specific to this stream. The material you hand the AI is your customers' personal data, and in a support conversation that can include sensitive details. Where and how that data is processed is a data-protection question from the start, not an afterthought, especially under European rules. That choice belongs in the plan next to the business case.

Measure the right thing. A resolution rate is easy to inflate, and a case marked solved because the customer gave up is not a win. Watch whether the customer came back satisfied, and whether the same issue returns, not just how many tickets the AI closed. This is the trap the Intercom numbers show, and it is easy to fall into when the dashboard rewards volume.

There is a quieter effect worth naming. When the AI takes all the easy cases, the people are left with only the hard and frustrated ones, which burns them out faster. And the easy cases were where new agents used to learn the ropes, so automating them away can dry up the pipeline of experienced people you will still need for the hard cases. A good rollout plans for both.

For Issue-to-Resolution the dependable end state is agentic with a human always one step away. Full autonomy is not recommended here. The Klarna walk-back is the evidence, and a customer's right to reach a person is close to a hard requirement, in practice and increasingly in regulation.

Figure 4: Gain and price across the four maturity levels. Each level brings a gain and asks a price, drawn from our cases. Agentic is the realistic target state, autonomous the warning case.

What comes next

Issue-to-Resolution shows the same pattern as the rest of the series, with a sharper edge: the routine work moves to the AI, the person moves to the cases that matter, and the whole thing depends on a gate, here the escalation to a human. Push past that gate and the savings turn into a trust problem.

The two headline cases here are the ones that carry the story most clearly, but the library holds more at each level. Salesforce and Commonwealth Bank sit alongside Klarna as agentic cases, and Air Canada, DPD and Cursor sit alongside the Klarna walk-back as autonomous cautionary cases, each with its own evidence and limits.