Three horizons of AI transformation.

The Three Horizons approach separates three different jobs. Horizon 1 improves the current company. Horizon 2 redesigns important workflows around AI. Horizon 3 builds an AI-native successor beside the current company. All three can run at the same time.

The original Three Horizons of Growth framework was developed to help leaders manage the current business while creating future options. Dreamwith adapts that logic to organisational AI transformation. This is our interpretation, not a claim that McKinsey endorses dreamwith or this service.

Horizon 1: improve the existing system.

Horizon 1 protects today’s customers, revenue and obligations. People use AI inside the company as it already operates. They may draft faster, search documents, summarise meetings, analyse information or automate repetitive fragments.

The goal is a healthier current company—not transformation theatre. Useful Horizon 1 work can reduce waste and improve decisions while preserving the process, software and roles that remain necessary.

Typical Horizon 1 work

• Give people safe access to appropriate AI tools.
• Improve document search and knowledge retrieval.
• Automate low-risk repetitive tasks.
• Strengthen data quality, security and governance.
• Measure where time, knowledge and margin are being lost.

Horizon 2: redesign the workflow.

Horizon 2 changes how a valuable piece of work moves. Agents can complete connected steps, handoffs can disappear and people can move from producing every intermediate task to directing and reviewing the system.

This is more demanding than installing a tool. The team must decide which source records are trusted, which agent owns each step, what counts as completion and when a person must intervene.

Typical Horizon 2 work

• Redesign a complete client-intake, proposal, research or delivery workflow.
• Give specialist agents bounded roles and shared context.
• Replace duplicated data entry and unnecessary handoffs.
• Build review, evidence and approval into the workflow.
• Change the tools and human roles around the new process.

Horizon 3: build the AI-native successor.

Horizon 3 asks a different question: if capable AI, shared company memory and coordinated agents had existed when this company began, how would we build it now?

The answer may require a separate operating architecture, product, team, cost structure and route to market. That is why dreamwith’s working thesis is to build the Horizon 3 company beside the current organisation. It can inherit the mission, expertise, customer problem and standard of care without automatically inheriting every meeting, application, handoff or reporting line.

Horizon 3 is not the old company after a better software rollout. It is the company we would build if AI existed from the start.

The horizons should run together.

The model is not a three-step maturity ladder. A company does not have to finish Horizon 1 before beginning Horizon 2 or Horizon 3. The horizons answer different questions and carry different risks.

Horizon 1 keeps the present healthy. Horizon 2 creates leverage and evidence. Horizon 3 protects the future from the constraints of the present. The mistake is forcing one team, one budget and one governance model to treat all three as the same kind of work.

Transfer only what has earned trust.

A separate successor is not an excuse to abandon the current company. It is a protected place to test a new operating model without placing current customers at unnecessary risk.

Work should move only after the successor demonstrates that it can meet the required standard. The transfer can be gradual: one workflow, one customer segment, one product or one operating capability at a time. Human owners remain responsible for the consequences.