Analytics and AI Solutions for Risk Based Inspections - Observatory of Public Sector Innovation

MoMRAH, is mandated to manage the Saudi cities, from regulation to operation, including commercial, construction activities, and municipal assets of streets, parks…etc That, translates into +50k km² of urban land, +550K streets, +500k commercial licenses… All that, is inspected with <7k inspectors! With challenges of productivity, prioritization, proactivity. Therefore, with sectoral platforms have immense data, comes a series of innovative AI solutions under the umbrella of "RiskBased" concept.

Innovation Summary

Innovation Overview

If we take the sheer size of the above-mentioned mandate and challenges, we realize the reflection on the massive operation in order to inspect, and if excellence is aimed, structural innovation is needed to change the rules of the game and to inspect based on risk and incentive the compliant. One of the innovative streams that MoMRAH has followed was the "Risk Based Inspection" using AI and ML to direct this massive operation, that is afterwards changed into a whole Analytics centre to carry on the culture of analysis and raising the leadership vision to deliver data-driven decisions.

In a closer prospective, MoMRAH introduced a mobile platform called Baladi, that opened the e-participation to report municipal requests and incidents, which through it has been receiving growing number of reports that follows the growth of the platform adoption reaching to +3mn annual cases in 2021. Half of that, +1.5 counts for a major national campaign to fight visual pollution, which puts enormous pressure on the operation arm of the ministry, municipalities. Hence, MoMRAH, represented by the Digital Transformation and Smart Cities Deputyship, formed a small team, later called "ClearVision", of internal and external resources, to target Analysis and Solutioning using Advance Analytics and AI to tackle this challenge.

Fast forwarding, the team focused on solving pains of Inspection, specifically, on Visual Pollution. Following a root cause analysis and a prioritization exercise of business pains, the team followed an agile 5-step methodology of: Ideate, Source, Develop and Test, Create Output, Deploy and Scale. For the below-detailed solutions, there were three main problems in inspection in order to optimize dispatching: no visibility of location-near and similar cases, prioritization is fixed and static, with almost absence of proactive inspection. Those problems were addressed with the following AI Engines, amongst other various CX and procedural solutions; Case Clustering, Case Prioritization, and Risk Based Dispatching.

Each one of them targets to tackle a problem;

Case Clustering: to cluster open cases that share proximity of location and classification.
Case Prioritization: takes +10 types of data sets, such as populations, points of interests, inputs from engine #1…and many others, to help score cases for importance and fasten response time to most important cases that requires a faster response or bigger population is exposed to.
Risk Based Dispatching: takes a similar input, but targets to dissect the city into grids and assign a risk score to each grid, to help pre-emptively identify visual-pollution-prone areas, in order to send inspectors, or enable crowd engagement of cooperative citizens.

These solutions, were novel to the sector, and different from any usual "software" development witnessed before. Moreover, in order to scale and implement throughout the value chain in order to integrate into the daily inspection activities, it took altogether around 9 months with a steep learning curve, and many changes to data quality, procedures, and even organization structure; adding new departments to own the ideation and another to host and operate the models, a third to handle the smart operations.

With each engine tackled a problem, all were weaved together, as the Clustering providing up to 30% savings in inspectors visits, Case Clustering showing the 20% of the cases that requires imminent response, and Risk Based Dispatching helps identifying 1200+ focused areas.

That said, that was the seed for embodying the Risk Based Inspection concept, followed by undergoing expanding the same tools into Retail, Health, Excavation, and other 4 types of inspection. Moreover, introducing new tools such as inspector demand forecasting, and more.

Innovation Description

What Makes Your Project Innovative?

Municipal operation in the kingdom always has been on static basis, or first in first out, never dynamic, or learning through historical cases or feedback loops coming from actual visits. Therefore, there was no much use of data, no data-based root cause analysis or solutioning.

Here comes the innovation we made around utilizing the plethora of data repositories MoMRAH sits on, where ClearVision provided;

Made use to many underutilized centralized data
Before providing a solution, provided a full scan of root cause analysis, recommendation, prioritization, and then arrived to solutioning
Provided solution to all participant municipalities, and a central mandate of inspection
Provided a dynamic view Vs the old static one
Fastened response and improved decision making
Helped improve data quality, since it relies on multiple types and sources of data, lots of feedback or demand were shared to data teams
Aimed to optimize operations and updates processes and procedures

What is the current status of your innovation?

The innovation has fully gone through the stages of development, from discovery and prioritizations, ideation and design, developing and testing, buy ins and business acceptance, implementation and full scale and process adaption. With a few months of reaping initial results benefits. Currently, the project is scaled up, and the innovation is in transition between next steps of full rollout and diffusing lessons and replicating the learning and expanding the concepts.

Innovation Development

Collaborations & Partnerships

Collaborators to this innovation were private sector, government officials, inspectors, and citizens. Each of them played a role, where private sector provided technical capabilities, government officials share problem and procedures and championing, inspectors shared the on-ground expertise, and the citizens were the main source of data and the main final beneficiaries.

Users, Stakeholders & Beneficiaries

Government officials (business owners) have a way better visibility on the incident's hot spots, priorities, and most important areas. Inspectors, have better optimization of their time and output. Citizens, receive a better response to their most important problems, and minimizing their need to reporting though pre-emptive capturing.

Innovation Reflections

Results, Outcomes & Impacts

Results of the innovation at hand is on the overall aspects of the operations of inspection, clustering is providing an average of 25% clustered cases that in great sense is saved trips. Also, the RBD was the only and main tool to help verify the results on an independent priority areas definition. On the other hand, Case Prioritization shows initial results where to identify the top quartile that the business wants to responds faster to, and the brackets are fully flexible.

Challenges and Failures

The learning curve coming with this innovation is steep. Out of which, it worth highlighting two main challenges; Roles and Responsibilities in the scale up, and structural challenges.

For roles and responsibilities, since this is solutioning from scratch utilizing new concepts, therefore, by nature is different from the current software development value chain, it was hard to have internal teams to own their usual part of this new entrant. Resistance was present, until we identified team members that have higher inclination to deal with such solution, and developed a clear RACI matrix.

For structural challenges, handling models, in terms of quality checking, maintenance, and daily operations was challenging since it requires different set of capabilities that are not present. Therefore, we used external parties to QC and verify the models with another third party, and now, we are creating an internal team for technical matters, and smart operations to handle the pipeline of models.

Conditions for Success

For such novel ideas we believe it requires; Pressing topics, championing, agility mindset.

First, Pressing Topics, from this experience, the visual pollution as a national priority to fix was the main driver for all the value chain from the problem to the adoption. Where it also helped to provide all the necessary needs for the project to see the light.
Second, Championing, a believing leader sets the bar high, and pave the way for problems arise throughout the way, aligns priorities when needed, align teams when misaligned, secure budgets when short is financials, and fasten things when bureaucracy blocks the way.
Third, agility mindset, to establish and improve, after identifying a business pain that is of a priority, then going through the 5-step development approach. That is, to create a version, that is better than random and matches/or better than human judgement, then release, and improve as you go.

Replication

The concept has started with the visual pollution related activities. Since this is now maturing, we are expanding the same concept to all other CRM and Inspection activities. In more details, we are:

Expanding the Case Clustering to all classification
Expanding the Case Prioritization to another 7 types of inspection
Expanding the Risk Based Dispatching to another 7 types of inspection

Which leaves us with more than 14 engines and we are introducing two master engines to work cross classification as well.

Lessons Learned

Three main lessons; Construct agile teams and assure openness, engage with clarity, and pave the role for industrialization. Creation of new concept can be frustrating for both teams and management; therefore, it is crucial to construct our teams well, and assure openness for discussing hurdles and support needed to make things happen.

Engage with clarity about responsibilities, once you have an MVP or so to have better planning and prioritization, as much as possible to minimize gaps of communication in such a cross-function innovation that is easily missed engaging critical teams, and by the time you reach to technical development or so, it would be quite late to engage.

Create with expansion in mind, lay procedure and transfer knowledge, to establish sustainability and growth, in this example, we laid the grounds for industrialization, now the new models, while more than double the number of engines, altogether, is taking almost half the time.