Case Study

An AI Knowledge Assistant for Energy Customer Service in Poland

Softwise.AI implemented an AI-powered knowledge assistant at innogy Polska S.A. - one of the largest energy companies in Europe, serving over one million customers across Polish households, SMEs, and global corporates - to support consultants in the call center and selected stationary branches in Warsaw. The platform unified knowledge from innogy's SharePoint, training documents, and external websites into a single natural-language search experience, letting consultants find accurate, current answers from one window - and at the Proof-of-Concept stage already delivered 84% response efficiency with 8.3/10 consultant satisfaction.

CLIENT

innogy Polska (now E.ON Polska)

YEAR

2024

INDUSTIRES

Energy & Utilities - Customer Service

DELIVERABLES

AI Knowledge Assistant for Consultants, SharePoint & External Source Integration, Inquiry Analytics, Knowledge Base Improvement Feedback Loop

The problem

innogy Polska is one of Europe's largest energy companies and, in Poland, serves over one million customers - households, small and medium-sized businesses, and global corporations. A customer base that broad creates a customer-service problem that is wide rather than deep: every consultant has to be ready to answer questions across the entire energy product range, under regulations that can change from one day to the next, drawing on documentation built and maintained by many different specialist teams.

In practice, that knowledge lived across SharePoint, training materials, and various innogy websites, with each team applying its own standards and no unified way to search across them. When a consultant couldn't immediately find an answer, the default was to ask a more experienced colleague or escalate to the second line - extending service time for the customer in front of them and, by extension, for everyone else those colleagues should have been helping. New consultants needed extensive training to be effective at all, and even tenured staff spent time hunting through documents to confirm whether what they remembered was still current.

innogy asked Softwise.AI to give consultants a single, intelligent way to search across the entire knowledge estate - and, just as importantly, to make the system itself a source of insight about where the underlying knowledge base needed work.

{Key Challenges}

Why a 1M+ customer energy provider couldn't scale knowledge across its call center?

The customer-service operation faced a structurally interlocking set of pressures:

Very wide knowledge surface per consultant
Every consultant had to cover the full breadth of energy products and customer types - from households to corporates - on a single hotline.
Frequently changing regulations and market conditions
Energy industry rules and pricing can change daily. Consultants had to constantly confirm whether the information they remembered was still current.
Slow, expensive onboarding and ongoing training
Getting a new consultant productive took weeks of preparatory training; veterans needed regular refreshers. The training overhead was a significant operational cost.
Heavy reliance on second-line support
A large share of inquiries was forwarded to more experienced colleagues or the second line - extending service times across the team.
A fragmented SharePoint estate
The knowledge base was maintained by multiple thematic teams without unified standards - making cross-team search and consistency hard to achieve, and slowing answer time on complex questions.

{Process}

How we deployed a one-window knowledge assistant and turned its usage data into KB improvements?

Softwise.AI delivered the engagement with a deliberate emphasis on closing the loop between the AI assistant and the underlying knowledge base.

1. Discovery and source mapping

The team worked with innogy to understand consultant workflow, the structure of the existing SharePoint estate, the training materials in use, and the external innogy websites that consultants needed to reference. The goal was a clear picture of what would need to flow into the single search window.

2. Source integration into one search surface

The platform was connected to SharePoint, training documents, and external innogy websites - bringing all the consultant-relevant knowledge into one place that could be queried in natural language.

3. Building the AI knowledge assistant

A natural-language search and retrieval layer was built on top of the unified estate, designed for the consultant's real workflow - quick lookups during a live call, with answers traceable back to source documents.

4. Proof-of-Concept with measurable success criteria

The system was deployed to consultants in the call center and selected Warsaw branches. Success was measured directly: 84% response efficiency and 8.3/10 user satisfaction were reached at the Proof-of-Concept stage.

5. Inquiry analytics turned into a KB improvement loop

The most valuable byproduct of the deployment was the analytics itself. By analyzing real consultant queries and the quality of answers returned, the team identified knowledge areas that were genuinely being asked about but were not described in the existing SharePoint documents - plus areas where the source content was simply out of date. That insight fed directly back into improving the SharePoint estate, refining periodic training, and shortening onboarding for new hires.

Solution

innogy's consultants now work through a single AI-powered search window that draws on the company's full knowledge estate - and gives the business a continuous, data-driven view of where its knowledge base needs to grow.

One window across SharePoint, training, and external sources

Consultants in the call center and Warsaw branches search SharePoint, training documents, and external innogy websites from a single interface - no more switching tools or remembering which team owns which document.

Natural-language search with traceable answers

Queries are made in natural language and matched semantically against the underlying content, returning the most relevant fragments with links back to the source - so consultants can confirm answers in seconds rather than skimming long documents.

Inquiry analytics as a knowledge-base lens

The system continuously analyzes what consultants are actually asking and how well the responses match. That data surfaces two things innogy could not see before: the topics being asked about that aren't covered in SharePoint, and the documents that have become outdated. Both go straight into the KB improvement backlog.

A feedback loop into training and onboarding

The same analytics drive better periodic training (focused on areas where consultants are struggling) and shorter onboarding (because the assistant carries part of the load that previously had to live in the new hire's head from day one).

{Results}

84% response efficiency and 8.3/10 user satisfaction at the Proof-of-Concept stage

The platform delivered measurable outcomes already during testing, before full production rollout:

84% response efficiency at the Proof-of-Concept stage.
8.3/10 user satisfaction from consultants using the platform.
Identification of knowledge gaps - topics being asked about that the existing documentation did not cover.
Identification of outdated content - SharePoint material that needed refreshing.
Better adjustment of external and internal communication to the actual needs of customers.

{Business impact}

Shorter service times, smarter training, and a self-improving knowledge base

Shorter customer service times. Consultants find answers from one window instead of hunting across SharePoint, training files, and websites.
Less dependency on second-line support. Fewer inquiries forwarded to more experienced colleagues means experienced staff can focus on genuinely complex cases.
Cheaper, faster onboarding. New consultants reach productivity sooner with the assistant alongside them, reducing the cost of high training overhead.
More targeted training. Periodic training programs adjust to the actual gaps the system surfaces - rather than the same generic curriculum year after year.
A knowledge base that improves itself. Inquiry analytics turn every consultant interaction into a signal about which documents need updating or extending - a continuous improvement loop the business now owns.

{Key recommendations}

5 principles for AI knowledge assistants in regulated, multi-team customer service

A few principles emerged from this engagement that apply to any large, regulated customer-service operation deploying AI:

Unify the sources before unifying the search. SharePoint estates maintained by many teams develop their own conventions and gaps. Pulling everything into one search surface is the first step; making the underlying material consistent is the second.
Always return the source. In regulated sectors like energy, where regulations change frequently, source-linked answers are how consultants confirm they're still current - and how the company defends a decision after the fact.
Treat inquiry analytics as a content backlog. The most valuable output of an AI knowledge assistant isn't the answer it gives the consultant - it's the list of questions it couldn't answer well. That list is the knowledge base's next quarter of work.
Use the assistant to redesign training, not just to answer questions. Real consultant usage data is a better guide to where training needs to focus than any internal opinion - and lets onboarding shrink in parallel.
Validate with measurable metrics from the PoC. Setting clear success criteria (e.g. response efficiency, user satisfaction) at PoC is what gives the business the confidence to scale - and the leverage to do so internally.

SUMMARY

Softwise.AI delivered a one-window AI knowledge assistant for innogy Polska's customer-service consultants, integrating SharePoint, training materials, and external innogy websites into a single natural-language search experience. At the Proof-of-Concept stage, the platform already achieved 84% response efficiency and 8.3/10 user satisfaction, while its inquiry analytics gave innogy a structured way to identify gaps in its underlying knowledge base, update outdated material, refine periodic training, and shorten onboarding for new consultants - turning a customer-service workflow improvement into a continuous, data-driven capability across the organisation.

TESTIMONIAL