Estha launches private AI workforce on Mac for business
Mon, 7th Sep 2026 (Today)
Estha has launched Estha on Mac, a private AI workforce for business users on Apple Silicon devices, aimed at small and medium-sized enterprises and professionals.
The Singapore-based startup says the software lets users run supported AI tasks on customer-controlled Mac computers instead of sending routine workloads to external AI services. It combines four pre-built business agents with no-code tools for creating private AI applications.
The launch reflects growing business interest in running more AI workloads on private infrastructure, particularly where cost control and data handling are concerns. Estha cited Broadcom research showing that 56% of enterprises were running or planning to run production AI inference on private cloud infrastructure, compared with 41% using public cloud for the same workloads.
Business tasks
Estha on Mac is designed around common office tasks. The initial set of agents covers meeting transcription and summarisation, document analysis, presentation creation, and spreadsheet analysis.
These tools, called Meeting Intelligence, Document Processing, PPT Maker, and Data Analyst, are intended for users without specialist AI skills. After installation and model download, they can be used without metered cloud inference charges.
For some workflows, the software can also operate without an internet connection. This allows supported confidential work to stay on the device instead of being routinely transferred to outside providers.
The approach may appeal to organisations handling sensitive records, including schools, financial advisers, legal firms, and healthcare providers. Potential uses include internal administration, contracts, client files, research materials, and operational documents.
AI adoption remains uneven among smaller businesses in Singapore. Estha pointed to IMDA figures showing that AI adoption among Singapore SMEs more than tripled in 2024 but still stood at 14.5%, compared with 62.5% among non-SMEs.
The company argues that the gap reflects practical barriers rather than lack of interest. Cost, data sensitivity, and limited in-house expertise have made many smaller organisations cautious about relying on public AI services or building their own systems.
Broadcom's survey also found that 62% of IT leaders were very or extremely concerned about the infrastructure costs of generative and agentic AI. Cost was the leading public-cloud concern among respondents, rising to 31% from 26% the previous year.
No-code tools
Estha is also offering two no-code tools, Quick Bot and Advanced Studio, for organisations that want to create their own private chat applications. These can be built from internal standard operating procedures, playbooks, and knowledge documents without writing code.
The resulting applications can support staff onboarding, training, internal advisory work, and access to internal knowledge. In each case, the applications run on the Mac rather than depending on an external AI stack.
Privacy is central to Estha's pitch. The company says local processing reduces the exposure associated with transferring sensitive information to external AI services, though organisations remain responsible for access controls, device security, policies, and human review.
"The work businesses most want AI to help with is often the work they are least comfortable uploading - the board paper, customer file, internal spreadsheet, contract or meeting recording," said Eric Lam, Co-founder and CEO, Estha.
"Estha on Mac brings AI to that work on hardware the customer controls, through applications employees can use without becoming AI engineers," added Lam.
Apple Silicon
The product is built for Apple Silicon and uses Apple's MLX framework and Metal. Estha says Apple Silicon's unified memory design allows the CPU and GPU to access a shared memory pool, reducing the need to duplicate model data between separate memory areas in local AI applications.
Metal handles compute workloads through the Mac's GPU. This technical design underpins Estha's focus on running supported AI workloads locally on widely used business hardware.
Customers are already using the software in education and advisory work, according to Estha. One example is United Christian College in Hong Kong.
"We've used Estha on Mac for at least three use cases. It allows us to build AI workflows involving sensitive data like students' grades, invigilation schedules without sending that information to an external AI service. The plug-and-play environment processes data locally - so tasks like analysing spreadsheets, documents and transcripts happen without confidential information ever leaving the device. Because users can also build and run their own AI applications locally, our teachers can use it without specialist AI knowledge," said Simon Lau, United Christian College (HK).
Another user, Credence Advisory, said the software had helped it handle internal queries and document work while keeping client information under tighter control.
"At Credence Advisory, we have always believed that clients deserve bespoke advice, not cookie-cutter answers. The challenge is that as the firm grows, more time is inevitably spent on repeated internal queries, document preparation and operational workflows. Estha for Mac gives us a way to scale our internal expertise while keeping confidential information under control. More importantly, it allows our senior people to recover time for the work that matters most - applying expert judgement to serve clients better," said Koay Tatt Chong, Director, Credence Advisory.