CX Impact — journey maps and funnels | Maksim Shelukhanov

CX Impact · Open-source tool for AI agents

Clear maps of customer journeys, service operations and sales funnels

Give your AI agent your documents, process descriptions and numbers. CX Impact helps turn them into visual maps, compare current and target scenarios, and prepare a shared reference for your team.

For product managers, analysts, designers and service teams.

Open source · For AI agents · Standalone HTML

Four CX Impact views: a customer journey, sales funnel, service blueprint and process using synthetic data.
Excerpts from actual CX Impact outputs. All examples use synthetic data.
01

One challenge, several useful views

01Customer journey map · CJM
What a person is trying to do, where they interact with the service, and which barriers your materials identify.
02Service blueprint
How customer actions connect to the work of staff, internal teams and systems.
03Process map
Who does the work, where responsibility changes hands, what conditions apply, and how the process recovers from a failure.
04AS IS / TO BE
What changes between the current and target scenarios, and what stays the same. Enable comparison when needed.
05Sales funnel
How many people reach each stage, how conversion changes, and where progress stops. Recovery and unresolved outcomes reflect the data you provide.

Requirements, observations, hypotheses and open questions remain distinct. The map helps your team agree on the next step and identify information that is still missing.

02

See the output

01

From product selection to order confirmation.

The journey map, service blueprint and process view describe the same task from different perspectives. AS IS / TO BE comparison helps your team review proposed changes.

AS IS / TO BE customer journey on synthetic data
02

Online sales funnel.

Stages from entry to a paid purchase, conversion between adjacent steps, recovery flows and a separate repeat-purchase calculation.

Sales funnel with synthetic cohort counts and definitions

These are teaching examples, not company results or a promise of higher conversion.

Want to explore the interactive views?

Download the example bundle, extract it and open index.html in your browser. It contains maps and funnels in English and Russian. Viewing the examples does not require an AI agent.

Download interactive examples · ZIP

Example bundle · 0.6.1

03

From materials to a map in three steps

  1. Install CX Impact in your AI agent. The GitHub guide covers installation and environment checks.
  2. Provide your materials and task. Describe the scenario and attach documents. For a funnel, include counts, the measurement period and participant-counting rules.
  3. Open and discuss the result. Review sources and unknowns, then share the HTML file. Supply the previous JSON with new information to create an updated bundle while keeping the original.
Read the installation guide
  • CX Impact renderer: Node.js 18+ to create and rebuild maps.
  • Verified Skills CLI 1.5.25 installer: Node.js 22.20+, npm and internet access.

Install with the Skills CLI

npx skills@latest add shelasmax/cx-impact --skill cx-impact

Manual installation

Download the CX Impact package · ZIP, extract it and copy the entire cx-impact/ directory, including SKILL.md, assets and renderer, to one of these locations in your project:

  • Codex and OpenCode: .agents/skills/cx-impact/
  • Claude Code: .claude/skills/cx-impact/
  • DeepSeek Harness: .agents/skills/cx-impact/ or .dsh/skills/cx-impact/

Back up any existing installation before replacing it, then start a fresh agent session after copying. Manual installation does not require Skills CLI; the renderer only needs Node.js 18+.

Choose your agent

The installer lets you select an agent. Your agent needs file and shell access to create and rebuild maps.

Agent Invocation Verification scope
Codex
$cx-impact
Tested local environment, installation and renderer.
Claude Code
/cx-impact
Documentation compatibility and installed package; no model-driven end-to-end test.
OpenCode
Ask to use cx-impact
Local skill discovery checked; no model-driven end-to-end test.
DeepSeek Harness
Ask to use cx-impact
File installation documented; harness execution not tested.

Other agents need file-based skill support and tools to run the renderer. Compatibility should be checked in the specific environment.

Example Codex request

$cx-impact
Use the attached materials to map the current customer journey and
service operations for placing an order. Preserve sources and distinguish
hypotheses and unknowns. Produce HTML for discussion and JSON for updates.
04

Share the result and build on it

You receive interactive HTML, editable JSON and a source snapshot for rebuilding. The viewer exports SVG; funnels also export a CSV table. Four designs and light or dark themes offer different presentation styles.

The HTML opens without a server or a CX Impact account. Creating a map with an agent uses your chosen model provider.

05

Turn a shared map into work on your business challenge

Maksim Shelukhanov

Maksim Shelukhanov is a business and transformation executive. My work connects e-commerce, omnichannel sales, customer experience, digital products and data.

You can use CX Impact independently. If you would like to work through a customer journey, process or points of drop-off together, we can discuss your challenge and the right engagement format.

06

Questions

Is this a separate online service?

It is a portable skill: a set of instructions and tools for an AI agent. Installation is documented for Codex, Claude Code, OpenCode and DeepSeek Harness. The resulting HTML opens separately in a browser.

What does it cost?

The project code is available under the MIT license. Your chosen AI agent and model usage depend on your plan. Consulting is a separate service.

Can I create maps without numbers?

Yes. Descriptions and documents work for customer journey and process maps. A quantitative funnel requires your data and metric definitions; missing values remain unknown.

Where do my materials go?

Map creation happens in your agent environment using the selected model provider. The generated HTML makes no network requests of its own. Follow your organization's rules when sharing materials with the agent.

Does the map identify the causes of drop-off automatically?

It presents the information you supply and the resulting calculations. Causes need to be checked against data and research; hypotheses and proposals do not automatically become facts.

Can I change the result?

Yes. Give the agent your previous JSON and new information, then save a separate updated bundle. The viewer does not include a visual editor.

07

Map your own challenge

Start with an example, then give your agent your materials.