Recommended resources
The freely available resources this lesson will direct you to.
Hello again. Last lesson, we separated Harborline’s reported problems from its proposed solution and untested assumptions. Now we can put those findings in order: what does a support agent do when a returns question requires policy guidance, and how might that work change?
By the end of this lesson, you will have two comparable maps for one target user: a current-state workflow that makes gaps in our knowledge visible, and a proposed workflow that responds to the reported pain points without pretending its design has been validated.
Start with one person and one task
A journey map tells the story of a person trying to accomplish a goal. Before drawing one, fix its boundaries. For Harborline, the actor is a support agent—not the sponsor, policy specialist, or customer. The scenario is a returns case that requires the agent to find guidance before replying. The journey starts when that need arises and ends when the agent either has guidance they can use or seeks help.
NNgroup’s short introduction identifies five useful ingredients: actor, scenario, phases, actions and mindset, and opportunities.
Watch “Customer Journey Mapping 101” from NNgroup for a compact model of what belongs on a journey map.
Watch the definition, then the five elements. Notice that the phases organize a story from the user's point of view; they are not a list of software components.
For our scenario, four phases are enough: recognize the information need, find guidance, assess it, and answer or escalate. We can use the same phases on both maps, making changes—and unresolved problems—easy to compare. A map need not record every click. It should reveal the transitions that matter, such as moving from a shared-drive search to a policy specialist when documents disagree.
NNgroup also distinguishes a map of today’s experience from a map of a possible future experience. Watch the distinction before we draft either one.
Journey Mapping: 2 Decisions to Make Before You Begin
In “Journey Mapping: 2 Decisions to Make Before You Begin,” NNgroup explains why current-state and future-state maps serve different purposes.
Watch the two states to distinguish documenting present friction from envisioning an improved journey. Then watch the blended approach: begin with a provisional current map, investigate it, and revise the proposed journey accordingly.
Map today’s journey without inventing evidence
The Harborline brief gives us fragments, not an observation of a complete case. It says agents search a shared drive, contact policy specialists when documents disagree, and sometimes copy outdated PDF text into replies. It does not tell us exactly how agents start a search, decide which document is current, or feel at each stage.
The responsible first draft is therefore a current-state assumption map. In the table, reported means stated in the client brief, not independently verified. Hypothesis means a plausible step or question to check with agents. The questions in the last column are our prompts for investigation, not quotations from users.
| Phase | Agent action and touchpoint | Friction or question | Evidence status |
|---|---|---|---|
| Recognize the need | Identify that a returns case requires guidance. | What in the case tells the agent they need to look something up? | Hypothesis: this is a sensible starting point, but the brief does not describe it. |
| Find guidance | Search the shared drive for returns guidance. | How does the agent narrow results or spot older copies? | Reported: agents search the drive; search behavior and result quality are unknown. |
| Assess guidance | Review what was found. If documents disagree, message a policy specialist. | Which guidance is approved for this case, and how is disagreement resolved? | Reported: conflicting documents prompt messages to specialists. The agent’s checking steps are hypotheses. |
| Answer or escalate | Use guidance in a customer reply, or wait for specialist help. | Could outdated text be used? How long does escalation take? | Reported: outdated PDF text is sometimes copied into replies. The timing and full escalation path are unknown. |
Read across a row to understand a phase, then down the action column to follow the journey. The important feature is the branch in “Assess guidance”: finding documents does not necessarily mean the agent can safely answer. A conflict can move the agent into a different touchpoint and involve another person. Similarly, an outdated PDF may look usable unless the agent has a reliable way to establish its status.
Do not draw an emotional curve simply because example journey maps have one. We have no agent testimony about frustration or confidence. A provisional mindset question—“Which document is approved?”—is useful because it directs research; assigning the agent a feeling would suggest evidence we do not have.
A workshop guide from NNgroup describes a practical way to improve such a draft: map what the team thinks happens, speak to people who do the task, change the map in light of their accounts, and identify the consequential pain points. Its future-state section then uses those pain points to shape proposed flows.
How to Run a Journey-Mapping Workshop: A Step-by- ...
Read NNgroup’s workshop guide for the method behind moving from an assumption map to a proposed journey. You do not need to run a full workshop to use that method here.
In “Part II: Current-State Mapping,” read Steps 1–4, from the description of making an assumption map through identifying pain points: current-state method. Focus on why the draft may contain gaps and why participants should describe their experiences before being shown the team's assumptions. In “Part III: Future-State Visioning,” read Steps 1–3, from generating ideas based on pain points to the consolidated group flow: future-state method. Focus on the connection between an observed difficulty and a proposed change, rather than on the workshop's voting mechanics.
For Harborline, we would seek a recent returns case and ask an agent to walk through what they searched, what they opened, how they judged the result, and what happened if sources conflicted. An example document pair and the associated specialist message could clarify more than a general opinion that “search is difficult.” We would update the map to reflect what actually happens—and preserve differences if agents take different paths.
Propose a better journey, including its failure paths
The future-state map uses the same actor, scenario, and endpoints. That makes it possible to ask whether the proposal changes the points of friction, rather than merely adding a chat screen. Here is a candidate journey for discussion, not a promise that Harborline’s policies, data, or systems already support it.
| Phase | Proposed agent experience | How it addresses today’s friction |
|---|---|---|
| Recognize the need | From the case-management screen, the agent asks for guidance about the returns situation, without entering sensitive case details unless that use is approved. | Keeps the lookup close to the task while respecting the security question left open in the brief. |
| Find guidance | An assistant presents relevant guidance the agent is permitted to access, with links back to the supporting documents. | Gives the agent a route to inspect sources rather than relying on an unsupported answer. Access rules still need confirmation. |
| Assess guidance | The agent opens the cited material and checks whether it applies. If sources conflict, are outdated, or cannot be established as approved, the assistant does not present a definitive policy answer. | Treats conflicting or uncertain evidence as a reason to pause, not as material to smooth over. |
| Answer or escalate | The agent uses checked guidance to compose the reply. If reliable, accessible guidance is unavailable, the agent follows a specialist escalation route. | Preserves human judgment and a path for cases the assistant cannot safely resolve. |
This is a user workflow, not yet a technical architecture. “Present relevant, permitted guidance” describes what the agent should experience; it does not specify how documents are ingested, retrieved, or access-checked. Those decisions come later in the course. Equally, “the agent checks” is a proposed behavior to test: we must learn whether links and document information actually help agents make that judgment.
The unsuccessful paths deserve space on the map. Suppose two apparently current policies disagree. The proposed journey should make that uncertainty visible and preserve a way to ask a specialist, rather than produce a confident synthesis. Suppose no document the agent may access supports an answer. The useful result is then an honest no supported answer and a clear next step—not a plausible-sounding reply. These branches are central to the experience, not edge cases to erase from an idealized diagram.
Notice also what the proposal does not claim. It does not assume the sponsor is right that citations alone create trust. It does not assume the lead’s 12-minute estimate is a measured baseline. It does not assume every old file is correctly marked, that the assistant can determine which policy wins, or that every escalation should disappear.
Use the two maps as a decision aid
A good pair of maps makes the reasoning between problem and proposal inspectable. For each proposed change, you should be able to point to the current phase it addresses: source links respond to uncertainty about guidance; a conflict route responds to disagreement between documents; an escalation route handles cases without a supported answer. If a proposed feature has no connection to the mapped task, ask why it is in this journey.
At the same time, a map is a research and design artifact, not proof of benefit. Before treating this proposal as settled, Harborline would need to check several things with agents and document owners: whether these are the real decision points in a case, how approved guidance is identified, what agents may access, and when specialist review is necessary. The maps help those conversations stay specific.
Keep the viewpoint clear. A policy specialist’s work of publishing updates matters greatly, but it is not a phase of the agent’s journey. We can mark “contact specialist” as an agent action and investigate what follows; mapping the specialist’s own work would require a separate viewpoint. This discipline prevents one diagram from becoming an unreadable mixture of user experience and internal operations.
Conclusion
To map a knowledge-retrieval workflow, choose one user and one bounded task, follow that user through meaningful phases, and record actions, touchpoints, questions, and important branches. Draft the current state with its evidence gaps exposed. Then propose a future state that addresses specific friction while showing what happens when guidance is conflicting, unavailable, or inaccessible.
Harborline now has two journeys it can put in front of agents and stakeholders for correction. In the next lesson, we will turn parts of the proposed journey into testable user stories—without losing the uncertainty and safeguards the maps revealed.