What comes next, and what we already know about it.
Generative AI is the current wave. Quantum computing is the next one. The wave after that is not yet visible. We read all three.
Every major shift in business technology arrives the same way. Decades of research, hardware acceleration, and patient capital. Then a year in which the technology appears to arrive overnight. Mobile did this. Cloud did this. Generative AI did this. The pattern is consistent enough to plan against.
Quantum computing is the next entry in the sequence. Google's Willow programme and IBM's quantum roadmap represent decades of investment that have not yet reached commercial use, and almost certainly will not for years. None of that means the trajectory is uncertain. The only open questions are timing and which categories of business problem will be reshaped first.
Artrilogic's position is that the bridge to the next era should be partially built before it has to be walked, and that the firm building it should be the firm already trusted with the current one. We are reading the trajectory now so the call we make in three years is the right one.
Research is not a service. It is a posture.
We do this work because customers commit to long engagements with our firm and deserve to know we are paying attention to what comes next. Not as a value-add. Not as a marketing surface. As part of being a senior partner.
Our customers bet on long-running engagements with us. They have the right to know we are still going to be the right firm in three and five years. This page is the evidence.
Four research streams, dated, and honest about the line between substance and hype.
Each stream below is reviewed every quarter. Where a stream connects to work we already deliver, the link is provided. Where it does not, it sits here and waits.
- Stream 01
Post-AI agent architectures
Why we watch. The current generative AI wave is the start, not the destination. What comes after the prompt era will reshape how businesses integrate intelligence into their systems.
Currently true
Agent orchestration through open standards (MCP, OpenAPI extensions for AI consumption) is moving from research to production. The firms building MCP servers over their existing systems today are buying themselves optionality for the next wave. This is one of the few areas where current investment has a clear forward payoff.
Currently hype
Most "agentic AI platform" marketing is rebadged workflow automation. Real agentic systems are rare in production and the ones that exist are narrow, well-bounded, and human-supervised.
Last reviewed: 10 May 2026 - Stream 02
Quantum computing, calmly
Why we watch. Decades of academic and industrial investment from Google (Willow), IBM, and others is converging on commercial viability. The same trajectory mobile and cloud and generative AI followed. The wave will arrive. The question is timing and what it actually changes for ordinary businesses.
Currently true
Post-quantum cryptography migration is the only quantum-adjacent decision most Australian businesses will need to make this decade. NIST has standardised PQC algorithms, ASD has issued guidance, and TLS stacks are starting to roll over. Any business with an integration estate should know whether their platform vendors have a PQC migration story. Beyond cryptography, useful business applications of quantum computing remain at least three to five years from production for the kinds of problems Artrilogic customers run.
Currently hype
"Quantum will transform every industry" claims, "Q-Day is imminent" panic narratives, and quantum-machine-learning marketing from firms with no working implementation. The genuine cryptographic implication is real. Most of the rest is twenty-year speculation dressed up as five-year urgency.
Last reviewed: 10 May 2026 - Stream 03
Sovereign AI infrastructure
Why we watch. Australian businesses are increasingly required, by regulator or by customer pressure, to keep sensitive data and inference inside Australian sovereignty boundaries. The shape of that infrastructure is still being worked out across the industry.
Currently true
On-prem and air-gapped inference is now technically feasible for a much wider range of model sizes than even twelve months ago. Sovereign LLM hosting in Australian data centres is moving from "rare" to "expected" for regulated customers. The firms whose AI architecture does not assume public LLM endpoints are positioned best for this shift.
Currently hype
"Sovereign AI" as a marketing label applied loosely to anything hosted in-region. True sovereignty includes data residency, model weights, inference logs, and fine-tuning provenance. Most vendor claims do not cover all four.
Last reviewed: 10 May 2026 - Stream 04
Open integration standards
Why we watch. Every wave of business technology is enabled or constrained by the integration standards underneath it. Generative AI is being shaped right now by MCP, by evolving OpenAPI extensions, and by the governance standards bodies are catching up on. The firms paying attention to the integration layer are positioned to ride the next wave. The firms ignoring it will be locked into whatever vendor wraps it best.
Currently true
Model Context Protocol is emerging as the default open standard for surfacing enterprise systems to AI consumers. Open integration platforms (WSO2, APISIX) are actively building MCP support. The gap between proprietary agent platforms and open agent platforms is widening, and the open side is winning on portability.
Currently hype
"AI-native integration platform" marketing from vendors who have added a chat box to a workflow tool. True AI-native integration means agent-shaped traffic patterns, semantic routing, and native MCP, not natural language wrappers on RPA.
Last reviewed: 10 May 2026
We do not believe research should change your roadmap.
Quantum is not a reason to abandon current investments. Generative AI is not a reason to panic-rebuild. The firm watching the horizon is the firm that lets you keep your current focus calmly.
We do not believe quantum computing should change a single decision your business makes this year, or probably next. We do believe it should be on the horizon of the firm you trust to advise you, so the call is right when it finally has to be made.
Bridges, built quietly, before you need to cross them.
Our bridges, not bonfires framing extends naturally to early-adoption decisions. Generative AI today, quantum-adjacent decisions tomorrow, something else after that. Our role is to make sure the bridge is partially built before you need to walk it. So when the call has to be made, it is a calm one rather than an anxious one, and the engineering work to get across has already begun.
This page is reviewed every quarter.
Each stream above carries a "last reviewed" date. If a stream has not been updated in a quarter, that is not because the page is stale. It is because nothing material has changed in our reading.
If you are reading this page and the most recent review date is more than four months old, please tell us. We have either lost the discipline or the page has broken.
Most recent review across all streams: 10 May 2026.
When something is ready to leave this page, we will tell you.
When a research stream matures into a capability we actually deliver, it leaves this page and becomes a service. Until then, it sits here.