The development of agents underpinned by data and knowledge assets looks set to be a major challenge for telcos. It may take up to ten years for telcos to truly master agents, but they will be key to the realisation of fully autonomous networks.

The journey towards an intelligence architecture providing distributed intelligence

A new generation of technologies, including agents, knowledge planes, reasoning engines and data mesh, are hailed as potential novel solutions to some of the most complex hurdles for telcos to reach Levels 4 and 5 in the TM Forum Autonomous Network Framework. We refer to these technologies collectively as an ‘intelligence architecture ’ with four dimensions, as shown in the figure below.

The four dimensions of an intelligence architecture

If you are not a subscriber, enter your details below to download an extract of the report

Achieving Level 4 and 5 autonomy will see telcos crossing a chasm of complexity from the earlier intelligence and automations of Level 3. There is a need to reimagine the management of the network with new intelligence and it is not clear whether the range of currently immature technical solutions will be able to support requirements. There are a number of intricately linked problems to be solved, including:

• Model capability
• Large models’ inaccuracy
• Not enough data
• Limited observability
• Model response too slow
• Issues inherent in all AI/ML (e.g., bias, hallucinations, drift)
• Explainability

Alongside a detailed overview of the challenges during the deployment of multi-agent systems (MAS), this report offers insightful recommendations to telcos seeking to embrace an intelligence architecture into their networks.

Table of contents

• Executive summary
  • A distributed intelligence is desirable but challenging to implement
  • Suggested early-day actions when developing an intelligence architecture strategy
• The journey towards an intelligence architecture providing distributed intelligence
• The challenges of creating distributed intelligence
  • The difficulties of building a MAS and supporting it with intelligence
  • Improving the intelligence of models to enable a MAS
  • Designing and running a MAS
  • Data and knowledge: the difficulties of supporting a MAS
  • Improving data and knowledge to support a MAS
• Summarising the barriers and the potential solutions to delivering a MAS
• Conclusion

Related research

• The journey to a self-healing network: Intelligence, agents and complexity
• Digital twins: Accelerating progress towards autonomous networks
• Achieving next-level network autonomy
• Finding value from AI, analytics and automation in the telco: Part 2

If you are not a subscriber, enter your details below to download an extract of the report

Charlotte Patrick

Charlotte Patrick

Associate Senior Analyst

Charlotte has 27 years of professional experience in strategy, marketing and finance. Most recently in the largest global technology analyst firm and previously two of the worlds largest global telecommunications companies. She is an electronics graduate and MBA with excellent business analysis, commercial and strategic planning skills.