For the last two decades, growth in our industry has been won on the fundamentals: module efficiency, financing structures, and the ability of an installer or EPC to design, price, and deliver a system faster than the next quote in the inbox. Artificial intelligence is now inserting itself into every one of those fundamentals at once. Not as a distant trend to monitor, but as a working part of the design software, procurement systems, and monitoring platforms our industry already touches daily.
System design
The most immediate impact is in design, and it's no longer theoretical. IBC SOLAR is rolling out OpenSolar's AI-powered auto-design tool, "Ada," for online estimates. The intent being that a prospective customer's site can be captured, modelled in 3D, and turned into a preliminary system design with an automatically generated bill of materials in seconds including estimated savings and an estimated PPA tariff before a site visit is required.
Platforms like Aurora Solar (AI roof capture and AutoDesigner) and SurgePV's Clara AI take this further, accepting natural-language design changes like "add a 25 kW carport, avoid the skylight" and re-running an 8,760-hour production simulation on the fly. What used to be a multi-hour proposal is increasingly a multi-minute one, with AI-generated layouts typically reaching 95 to 98% of the energy yield a manual optimisation would achieve. The honest caveat, is that these designs still need a human to review cases including shading obstructions, structural quirks and cluttered rooftops before they're locked in.
For BESS sizing specifically, AI models are improving how we match storage capacity to load profiles, tariff structures (including South Africa's time-of-use and wheeling economics), and backup requirements. This is moving sizing away from rules of thumb toward something closer to a genuine optimisation problem.
For inverters, AI is embedded at the control layer, and this is where several of our own OEM partners are already leading. Huawei's FusionSolar 9.0 platform, launched this year, pairs its grid-forming SUN2000-506KTL inverter with AI-driven O&M across the full plant lifecycle, positioning solar assets as active, grid-supporting infrastructure rather than passive generation. Sigenergy's Sigen Cloud platform runs AI-driven fault detection and predictive diagnostics across residential, C&I, and utility inverters and its SigenStor battery range, and on a manufacturing case study it cited a 13% net revenue uplift from AI-driven dispatch versus static operating modes. Victron Energy's VRM portal, a platform IBC SOLAR installers know well, combines an AI model of a site's solar production potential with satellite-based irradiance forecasting to generate day-ahead and six-day yield predictions, alongside automated alarms that flag potential issues before a technician needs to visit. SolarEdge's HD-Wave MPPT and Enphase's IQ platform apply similar logic at module and inverter level, dynamically adjusting output in response to shading, weather, and load in real time. This is the difference between a system that simply generates power and one that actively manages when and how that power is used, stored, or exported.
What this means for installers, EPCs, and the sales function
This is the part of the conversation that matters most to the people actually closing deals. AI does not replace the relationship-driven, trust-based nature of solar sales. A system on someone's roof or factory floor is still a significant capital decision, and clients want to deal with a credible sales professional.
What AI changes is the workload around that relationship:
- Turnaround time.A proposal that used to take a day can be produced, priced, and simulated in an afternoon, which compresses the sales cycle and means the installer who responds fastest with a credible, accurate quote increasingly wins the deal.
- System costs.Faster design cycles and better-optimised BESS/PV sizing reduce engineering overhead and can shave real percentage points off installed cost. These savings flow through to a more competitive quote.
- Scale without headcount.Smaller installers can now produce a volume of accurate proposals that previously required a design team, levelling the competitive field between large EPCs and smaller regional players.
- Differentiation shifts.As design and proposal generation commoditise, the sales edge moves toward financing knowledge, after-sales service, and genuine technical credibility which are the things AI cannot easily replicate.
For distributors like IBC SOLAR, this means our value to installer partners increasingly lies not just in product, but in the tools, financing relationships, and technical support we can bundle alongside it.
Purchasing, logistics, and stock management
This is where I think the underappreciated value sits. AI-driven demand forecasting platforms, the same category of tool used across broader supply chains can materially improve how a distributor times purchase orders against shipping lead times, currency movements, and known project pipelines, reducing the carrying cost of stock that would otherwise sit on the balance sheet losing value as newer, cheaper product generations arrive. OpenSolar's platform is a good illustration of this trend reaching the installer level directly. OpenSolar's integrated hardware ordering pulls straight from the design and bill of materials, with distributor account management built into the same workflow. Logistics optimisation which includes container routing, customs documentation, warehouse allocation is a less glamorous application, but for a market like South Africa, where freight and import lead times are a real constraint on project timelines, this is where AI can quietly remove weeks from a project's critical path.
Maintenance
Predictive maintenance is arguably the most mature AI application in our industry today, and it's well represented across our own supplier base. Sigenergy's Sigen Cloud combines proactive fault detection, automated root-cause analysis, and an AI-powered support chatbot to compress response times across a fleet of devices, including third-party inverters and diesel generators, not just its own hardware. Huawei's FusionSolar applies AI-driven O&M across its Smart PV stations to reduce truck rolls and extend plant lifecycle value. On the monitoring and portfolio-management side, meteocontrol's VCOM platform, used across large PV portfolios globally, layers machine-learning-optimised forecasting on top of measurement data and satellite imagery to flag underperformance early, with its Digital Twin feature visually pinpointing string- or module-level anomalies faster than table-based monitoring alone; it's proven its worth at scale, correctly tracking a 13 GW drop and recovery in German solar output during a recent solar eclipse. Victron's VRM applies the same forecasting logic at the site level, letting installers move from reactive site visits to proactive service contracts. On the storage side, SolarMD, a South African-manufactured lithium battery producer, designs its battery management system in-house for cell-level monitoring and protection, reflecting the same shift toward smarter, data-driven storage management that's becoming standard across the industry.
For O&M providers and EPCs carrying multi-year service agreements, all of this shifts the economics of maintenance from a reactive cost centre to a manageable, forecastable one, and it materially improves the client's actual realised yield, which is ultimately what determines whether they refer the next customer.
The economic case
Taken together; faster design, tighter procurement, reduced maintenance cost, and shorter sales cycles, AI's economic contribution to PV is a genuine reduction in the cost of getting a system from quote to commissioned asset. In a market like ours, where the C&I segment is the primary growth engine and margins are under real pressure from module price volatility, that efficiency gain is not optional. It is becoming table stakes for staying competitive.
Project Finance and Funding
One area worth a closer look, distinct from the utility-scale project finance conversation, is what AI is doing to funding for commercial and industrial rooftop and ground-mount projects. For funders, AI-assisted underwriting is starting to compress diligence timelines that have traditionally been a drag on deal speed. This could include automated analysis of an offtaker's payment history, site load profiles, and PPA structure can flag counterparty and offtake risk faster than a manual credit review, and AI-generated yield forecasts (the same category of tool referenced earlier for design and O&M) increasingly feed directly into a funder's own DSCR modelling rather than requiring a separate, generic engineering report.
For a market with 20-plus alternative funders and C&I IPPs actively competing for deal flow, this speed advantage is becoming a genuine point of differentiation between funding partners. For installers and EPCs, the practical upside is a shorter gap between a signed proposal and a funded, ready-to-build project. This is critical in a segment where deals are often won or lost on how quickly a credible funding offer can be put in front of the client.
The risk, however, sits on both sides of the deal. Funders leaning too heavily on automated risk scoring may underweight site-specific realities an experienced credit analyst would catch (a smaller offtaker's genuine payment reliability doesn't always show up cleanly in the data), while installers relying on AI-generated production estimates to pitch financing terms need those numbers to be defensible enough to survive a funder's own underwriting, not just fast to produce.
The risks
Cybersecurity is the most serious risk. Solar inverters and BESS units are increasingly connected, remotely manageable devices. This connectivity is precisely what AI-driven monitoring and control depends on. For an industry built on distributed, internet-connected assets, a breach is no longer a theoretical IT problem. Any organisation deploying AI-driven monitoring or control needs to treat this as a genuine operational risk, not a footnote.
Over-reliance on automated design. AI-generated layouts are good, but not infallible. A design that hits 95–98% of optimal yield on paper can still miss site-specific realities like a shading obstruction, a structural constraint, a local grid connection quirk that an experienced engineer would catch. Best practice remains AI-assisted, human-reviewed design, not AI-only.
Commoditisation pressure on margins. If every competitor can generate an accurate, well-priced proposal in minutes, the differentiation that used to justify a premium quote erodes. This is good for clients and challenging for installers who haven't built a value proposition beyond the design itself.
Data and IP exposure. Client site data, pricing models, and project pipelines increasingly flow through third-party AI platforms. Organisations need to understand exactly what data these tools retain and how it's used before adopting them at scale.
Workforce disruption. Design and proposal roles will change in nature, not necessarily disappear — but organisations need to be honest with their teams about that transition rather than pretending it isn't happening.
Cybersecurity interventions installers and EPCs should be planning for
- Adopt IEC 62443 as the working framework.Originally written for industrial control systems, it's rapidly becoming the reference standard for solar and DER cybersecurity.
- Network segmentation and zero-trust access.Keep inverter and BESS communications on a separate network segment from general business IT, with strict access controls on who, human or system, can reach that segment.
- Firmware discipline.Insist on vendors that support authenticated, auditable remote firmware updates, and build patching into the maintenance contract rather than treating it as optional.
- Vet the supply chain, not just the product.Cybersecurity due diligence now needs to extend to component sourcing, not just software — a relevant consideration when specifying OEM partners.
- Monitor for anomalies, not just faults!
Where this leaves us
AI is not going to sell a solar system on its own, and it is not going to replace the trust an installer builds with a client over a site visit and a well-explained proposal.
It is already compressing the time, cost, and design overhead that used to sit between a lead and a signed contract — and the organisations that treat AI as core infrastructure, with the security discipline to match, are the ones that will hold their margins as the rest of the market catches up.
I'd welcome others' views, particularly from installers and EPCs in the South African market.
In keeping with the subject of this article, AI played a significant role in drafting it. The ideas, industry perspective and final review are mine, but much of the initial writing was AI-assisted. Perhaps that's the most practical illustration of the point I've been making: AI doesn't remove the need for human expertise. It changes how quickly we can turn that expertise into something useful. Another example of how AI is impacting marketing and content creation in our sector (if you view this article as such).