AI governance push hits barriers
The Asian Development Bank (ADB) has identified that Pakistan's push to modernise public services through artificial intelligence is confronting deep structural barriers, as fragmented data systems, weak inter-agency coordination, low digital literacy and limited institutional capacity continue to constrain access to government services and tax administration, raising concerns that technology-driven reforms may outpace the state's ability to implement them effectively.
These challenges were identified by Farzana Noshab, Lead Economics Officer at the Public Sector Management and Governance Sector Office of the ADB Sectors Group, and Yuji Miyaki, Public Management Specialist (Taxation) at the ADB Sectors Department, in a blog post published on the ADB website.
"Improving government efficiency through AI depends less on developing new algorithms or complex models than on allowing public officials to design the tools they use," they said.
The adoption of AI in public sector entities, along with the contribution of local players, including the input of government officers in developing AI-based solutions, is greatly needed in Pakistan to replace complicated and time-consuming processes, said Mehwish Salman Ali, CEO and Founder of Data Vault Pakistan and ZahanatAI.
She pointed out that local AI tools and chatbots are being encouraged in public sector departments, particularly to structure databases at a faster pace and facilitate the public in service delivery. However, she stressed that the integration of public and business data should remain within the country, housed in local data centres.
In Pakistan, attempts to modernise public services often rely on external suppliers who design systems without fully understanding the needs of agencies or citizens. This approach can lead to slow adoption, limited ownership and systems that do not align with on-the-ground realities.
Instead of receiving pre-built solutions, government teams should define their needs and help shape system functions from the start. This approach allows artificial intelligence to be introduced where it delivers the most value, the blog noted.
For instance, the Federal Board of Revenue has recently streamlined tax administration and filing procedures. These reforms include launching a user-friendly electronic tax return form with auto-fill features and developing tailored forms for different taxpayer categories, including an upcoming Urdu version. The online filing system has also been upgraded for real-time data sharing, while withholding tax payments have been automated and expanded to include digital invoicing. These efforts reduced previously complex and numerous requirements to a limited number of straightforward steps. This streamlining not only made compliance easier for citizens but also improved transparency and efficiency for officials, the blog said.
By eliminating unnecessary steps, standardising documentation and harmonising data formats across agencies, digital tools and AI functioned more effectively. The result is a tax and customs administration that is more accessible, less prone to errors and more transparent in serving the public, the ADB noted.
Improving government efficiency through AI depends less on developing new algorithms or complex models than on allowing public officials to design the tools they use. In line with this, tax officers and information technology staff in Pakistan worked together to develop an AI-enabled tax analytics module and a multilingual citizen services chatbot.
Introducing AI into government systems requires high-quality data, consistent formats and clear rules on data sharing. At present, databases across agencies are not always compatible, reducing the accuracy of AI tools. A multilingual AI-powered chatbot aims to simplify information, guide users through government services and offer audio or video explanations for people with limited literacy or internet skills.
The chatbot must therefore be designed to support accessibility rather than increase inequality. Over time, tools such as the chatbot could expand into other services, including employment, health and disaster response. Early planning for these expansions can support long-term integration. If Pakistan does not address the complexity of its public services, many citizens will remain excluded.
Dr Noman Ahmad Said, CEO of SI Global Solutions, said AI adoption and its desired outcomes will only be realised once a significant segment of the population, including government officials, learns the efficient use of AI tools and solutions.
Immediate steps are needed to provide training to the public, develop online courses and design boot camps at universities and public sector institutions to achieve early and sustainable outcomes, he said.